feat(opstack): ClassifyTile consumes ClassifyBox for cave archetypes -- the T1.d prize, behind the same opt-in

⚠️ THIS IS THE ONE COMMIT OF THIS RUN WHOSE FAILURE MODE IS A HOLE, NOT A REGRESSION. A non-Mixed
verdict makes the world skip GenerateMesh entirely: no triangles, no collision, invisible until a
player falls through. Read VoxelForge.OpStack.ClassifyTileSoundness before trusting it.

WHAT CHANGED
ClassifyTile used to `return Mixed` without a call for every cave archetype ("archétype cave [...]
pas prouvable en v1"). It now builds that strate's stack and folds ClassifyBox. SurfaceWorld and
bedrock gaps keep their hand-written proofs untouched -- the exact-lattice column test is strictly
better than any box bound, so the stack has nothing to add there.

ONE DEFINITION OF THE STACK MAPPING, and that is the load-bearing part
GetDensityAt's ~90-line build switch is extracted into VF_BuildOpStackForChunk and both callers now
use it. A second copy would be the worst bug available here: a tile skipped on the verdict of a
stack that is not the one producing its density is precisely a hole. A "keep these in sync" comment
would not have been enough; there had to be only one. Params are passed in by pointer, never fetched
inside, because both callers already have them.

SIX GUARDS, ALL FAILING TO MIXED, none giving the benefit of the doubt
1. The strate must actually opt in -- on EVERY chunk coord the box touches, not just the one that
   triggered the attempt. Otherwise the classifier judges a field the mesher will not produce.
2. One cave slot per tile. Two slots means two param sets, and a stack answers only for its strate.
3. No mixed cave/surface/gap tile: the cave stack's box would cover Z belonging to another strate.
4. The params must be BIT-IDENTICAL across every chunk coord the box touches. This is the guard that
   matters, and it exists because of the finding committed earlier today: GetGenerationParams blends
   params inside a strate (Alpha depends on chunk Z for Gradient, and on chunk XY as well for
   Interleaved), so one stack genuinely cannot represent a tile that straddles a transition band.
   Memcmp on POD: differing padding can only produce a false MISMATCH, i.e. one Mixed too many.
5. A 27-chunk-coord cap, so a very wide tile does not pay for the check. We give up the gain, never
   the safety.
6. The disturbances are folded by hand (chasm ⇒ CarveOnly, bridge/ridge ⇒ FillOnly), because
   DECOMPOSITION 10.2 leaves them OUTSIDE the stack -- GetDensityAt applies them after the negate.
   A verdict that ignored them would be wrong exactly where they act. Same inequalities the
   SurfaceWorld branch already uses.

The diff layer needs no new guard: ClassifyTile already returns Mixed for any tile with player mods
in range, before any of this.

TEST: VoxelForge.OpStack.ClassifyTileSoundness -- the same brute-force oracle as the existing
ClassifyTileSoundness, on a world where every strate opted in. It does not check the fold (that is
BoxVerdictFold) or the operators (those are the eight equivalence tests); it checks the WIRING. The
number to read first is the count of tiles actually brute-forced: zero non-Mixed verdicts would mean
the test proved nothing, so that case is an ERROR rather than a quiet pass. Its failure message
lists the four suspects in the order worth checking.

FIXTURE: FTestWorld::Build gains a bUseOperatorStack parameter (default false, so the thirteen
existing tests still exercise the switch), and every test world now gets a PROCESS-UNIQUE
LayoutVersion. That second change fixes a real cross-test hazard that was only ever hidden by an
accident: PassagesVersion is per-instance and starts at 0, so two FTestWorlds both reported 1, and
GetDensityAt's per-chunk caches are keyed on (ChunkCoord, LayoutVersion) -- one world could be
served the previous world's params AND its CP_UseOpStack flag. Invisible while every world agreed
the flag was false. The first world that ticks it removes that coincidence, in both directions.

WHAT THIS BUYS TODAY: Maze, FlatPlain/CrystalChamber, VerticalShafts and FloatingIslands can now
prove tiles in production, which is where the measured skipping (Maze 23/60, slabs 36-40/60) turns
into frames. TunnelNetwork and Underwater still prove nothing: their chain dies at FRoomGraphSource,
which answers Both with unknown amplitude. Making it answer spatially means building the SDF cache
for the queried box -- now worth doing, since a skipped tile saves 30k+ density evaluations, and the
amplitude fold plus the modifiers' Identity inheritance are already in place to receive it. That is
the next piece.

And nothing here changes Jahni's current world: no strate asset has bUseOperatorStack ticked, so
every guard above is unreachable in his project until he ticks one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

CODEMAP.md

@@ -124,8 +124,15 @@ in its per-chunk refetch block and evaluates it *instead of* the `switch` only w
 archetype in the ported list, which is now **all 8 of 8**: Maze, FlatPlain, CrystalChamber,
 SurfaceWorld, VerticalShafts, FloatingIslands, TunnelNetwork, Underwater. A strate that has not
 ticked the box still takes the `switch`, unchanged — the flag is the only thing that switches paths.
-**`ClassifyTile` is NOT wired** — still hand-written guards, not `ClassifyBox`. That is the next
-piece, and it is where the measured tile-skipping turns into frames.
+**`ClassifyTile` IS wired now**, for CAVE archetypes only and behind the same per-strate opt-in:
+where it used to `return Mixed` without a call, it builds the strate's stack through the *same*
+factory `GetDensityAt` uses (`VF_BuildOpStackForChunk`) and folds `ClassifyBox`. SurfaceWorld and
+bedrock gaps keep their hand-written proofs — the exact-lattice column test is better than any box
+bound. Guards, all failing to `Mixed`: one cave slot per tile, no mixed cave/surface/gap tile, the
+opt-in true on *every* chunk the box touches, the params **bit-identical** across every chunk coord
+the box touches (blended transition bands make one stack unable to represent the tile — `AUDIT §C2`),
+a 27-chunk-coord cap, and the disturbances folded in by hand since they are applied after the stack.
+Brute-forced end to end by `VoxelForge.OpStack.ClassifyTileSoundness`.
 ⛔ Never run both paths in one world. **Comparing them IS legitimate now** — the ~1 ULP residue of
 AUDIT §C10 is gone since `FPSemantics = Precise`, and all eight equivalence tests compare bit for
 bit. They are port-correctness oracles, not fidelity checks: the acceptance bar is §2.6.1 (same seed
@@ -259,7 +266,8 @@ redesign; tile identity lives in `FVoxelTileKey` (VoxelWorld.h).
 | **`GetSlabDensity`** | 1306 | FlatPlain/CrystalChamber pipeline. See §4.2. |
 | `SampleSurfaceStructuralZ` | — | **F20:** the RAW SurfaceWorld heightfield (continents+mountains+detail), BEFORE any terrain op; returns terrain Z + relief M. Cliff re-samples it at an XY offset for a cheap analytic slope. |
 | `ComputeSurfaceTerrainZ` / `GetSurfaceDensity` | — | SurfaceWorld heightfield → terrain Z, then density; biome **output-blend** lerps dominant/neighbour heights (`ParamsD`/`ParamsN`/weight). **F20 surface ops** (`FSurfaceGenerationParams`, biome-selected + slope/relief-conditioned, all default off): Cliff (slope-gated STEEPENING — push height from local mean where steep ⇒ sheer walls; 4 structural resamples only when on), Terrace (relief-gated + `TerraceHardness`), LayerLines (sedimentary shelves) — pure per-column height REMAPS applied here so the single height oracle stays consistent (MC/sheets/ClassifyTile/deco/BP bridge). **Phase 2 OVERHANG** (volumetric — real jutting shelves): in `SurfaceDensityFromColumn`, for AIR voxels in a window `(TerrainZ, TerrainZ+OverhangHeight]` above a steep slope, the heightfield is re-sampled UPHILL (toward the cliff) by a reach that GROWS with height (tiny low ⇒ air over the void, full high ⇒ borrows the far cliff rock) and unioned in ⇒ a shelf attached to the cliff, tapering out over the void with air beneath (the sketch). Per-column `OverhangAmp`(=strength·slope-gate) + unit uphill `(DirX,DirY)` resolved once in `ComputeSurfaceColumn` (gradient sampled at the REACH scale so a spot over the void can see the cliff), cached on `FSurfaceColumn`. Genuine 3D (per-voxel structural re-eval, gated to steep overhang columns). Off ⇒ byte-identical. §8.14. |
-| `ClassifyTile` | — | **T1.d trivial-tile reject** (worker, called by `LoadTile` before `GenerateMesh`): proves a tile AllSolid/AllAir on the mesher's exact lattice (gap chunks + SurfaceWorld columns via the SHARED `GSurfColCache`; seal bands; guards: diff mods, passages, spine, disturbances, **F20 overhang** — a column point in `(TerrainZ, TerrainZ+OverhangMargin]` (margin = max `OverhangHeight`) is unprovable ⇒ Mixed, UPWARD only since the shelf union only ADDS rock above ground, so an overhang shelf never holes a trivially-skipped tile) → skip gen. Mixed = generate normally. §8.10. |
+| `VF_BuildOpStackForChunk` (file-static) | — | **The archetype → stack mapping, written down once.** `GetDensityAt` and `ClassifyTile` both call it; params are passed in, never fetched here. A second copy would be the worst bug available in this file — a tile skipped on the verdict of a stack that is not the one producing its density is a hole. Returns false (⇒ caller falls back to the `switch`) for an unported archetype, missing params, or a **degenerate strate**, since five archetype functions early-out to air there and the stack deliberately has no such early-out. `Refs.Surface == nullptr` makes it refuse SurfaceWorld, which is how `ClassifyTile` keeps its own exact-lattice proof. |
+| `ClassifyTile` | — | **T1.d trivial-tile reject** (worker, called by `LoadTile` before `GenerateMesh`): proves a tile AllSolid/AllAir on the mesher's exact lattice (gap chunks + SurfaceWorld columns via the SHARED `GSurfColCache`; seal bands; **cave archetypes via `FVoxelOpStack::ClassifyBox` when the strate opted in** — see §3.2d for the six guards, all failing to `Mixed`; guards: diff mods, passages, spine, disturbances, **F20 overhang** — a column point in `(TerrainZ, TerrainZ+OverhangMargin]` (margin = max `OverhangHeight`) is unprovable ⇒ Mixed, UPWARD only since the shelf union only ADDS rock above ground, so an overhang shelf never holes a trivially-skipped tile) → skip gen. Mixed = generate normally. §8.10. |
 | `SampleRelief` / `SampleMoisture` | — | Climate fields (pure XY, [0,1]). Relief = shared source of truth for the relief map M. §8.14. |
 | `SampleBiomeAt` | — | Warped-Voronoi + climate biome query (dominant + neighbour + weight). Reference used by the preview bake + `GetDominantBiomeAt`. §8.14. |
 | `ResolveBiomeSampleAt` / `RebuildBiomeGrid` | — | Hot-path biome resolve (FBiomeSample) via a box-validated per-chunk cell-grid cache. Bit-identical to `SampleBiomeAt`. §8.14, §8.10. |