streaming: deeper ready queue + startup apply cap; tile-local mouth-support culling (L72)

- Unreal: ready queue depth 8 -> 16 (cap 4 worker-sets / 64), startup mesh-apply cap
  12 -> 24 within the unchanged 14 ms budget. Owner SW arrival, same 2,850 keys:
  initial set ready 4.78 -> 3.74 s, queue p95 123 -> 35 ms, request-to-apply p95
  139 -> 45 ms; walkable gate 1.07 -> 1.11 s; per-frame apply p95 0.73 ms.
- Core composer: mouth-support segments filtered by tile bounds, mouth relief result reused
  (route checks -80/-87% in passage tiles). Hashes identical W1/W8/W12, Clang and GCC.
- Opt-in per-tile composer counters. Verify control meshes stored as .xz (78.5 -> 18.4 MB).
- Fast tier 269 PASS.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1

Source/VoxelForge/Private/VoxelWorld.cpp

@@ -238,11 +238,11 @@ namespace
         GVoxelForgeStreamingResultBacklogCap,
         TEXT("Maximum queued results plus admitted tile tasks; 0 disables the safety cap. This is separate from the worker-slot cap."));
 
-    int32 GVoxelForgeStreamingReadyQueueDepth = 8;
+    int32 GVoxelForgeStreamingReadyQueueDepth = 16;
     FAutoConsoleVariableRef CVarVoxelForgeStreamingReadyQueueDepth(
         TEXT("voxel.StreamingReadyQueueDepth"),
         GVoxelForgeStreamingReadyQueueDepth,
-        TEXT("Extra admitted tile tasks kept ready beyond the worker slots, up to one worker-set; 0 disables the refill queue."));
+        TEXT("Extra admitted tile tasks kept ready beyond worker slots, up to four worker-sets/64 tasks; 0 disables the refill queue."));
 
     int32 GVoxelForgeStreamingBackgroundWorkerCap = 1;
     FAutoConsoleVariableRef CVarVoxelForgeStreamingBackgroundWorkerCap(
@@ -312,8 +312,8 @@ namespace
     FAutoConsoleVariableRef CVarVoxelForgeStreamingStartupApplyBoost(
         TEXT("voxel.StreamingStartupApplyBoost"),
         GVoxelForgeStreamingStartupApplyBoost,
-        TEXT("Bounded cold-view upload budget (default on): 1 allows a 14 ms / 12-mesh budget until the initial desired set and walkable ring are ready; 0 uses the authored per-frame budget."));
-    constexpr int32 StreamingStartupMaxMeshApplies = 12;
+        TEXT("Bounded cold-view upload budget (default on): 1 allows a 14 ms / 24-mesh cap until the initial desired set and walkable ring are ready; 0 uses the authored per-frame budget."));
+    constexpr int32 StreamingStartupMaxMeshApplies = 24;
     constexpr float StreamingStartupApplyBudgetMilliseconds = 14.0f;
     constexpr double StreamingStartupApplyMaxSeconds = 30.0;
 
@@ -2207,7 +2207,7 @@ void AVoxelWorld::BeginPlay()
                       CommandLineStreamingReadyQueueDepth))
     {
         GVoxelForgeStreamingReadyQueueDepth = FMath::Clamp(
-            CommandLineStreamingReadyQueueDepth, 0, 32);
+            CommandLineStreamingReadyQueueDepth, 0, 64);
     }
     if (FParse::Value(FCommandLine::Get(), TEXT("voxel.StreamingBackgroundWorkerCap="),
                       CommandLineStreamingBackgroundWorkerCap))
@@ -2634,7 +2634,8 @@ void AVoxelWorld::BeginPlay()
         FTaskGraphInterface& TaskGraph = FTaskGraphInterface::Get();
         LowLevelTasks::FScheduler& TaskScheduler = LowLevelTasks::FScheduler::Get();
         const int32 QueueDepth = FMath::Clamp(
-            GVoxelForgeStreamingReadyQueueDepth, 0, GetMaxConcurrentStreamingTasks());
+            GVoxelForgeStreamingReadyQueueDepth, 0,
+            FMath::Min(64, FMath::Max(1, GetMaxConcurrentStreamingTasks()) * 4));
         VoxelForgeStartupTrace::RecordEvent(TEXT("worker_backend_config"), FString::Printf(
             TEXT("\"logical_cores\":%d,\"effective_tile_slots\":%d,"
                  "\"task_api\":\"UE::Tasks::Launch\",\"task_scheduler_workers\":%u,"
@@ -11873,11 +11874,10 @@ void AVoxelWorld::UpdateChunksAroundPosition(const FVector& CenterPosition, APaw
         }
     }
     const int32 MaxTasks = GetMaxConcurrentStreamingTasks();
-    // Keep at most one worker-set of ready work so an admission pass can feed idle slots
-    // without letting far work consume the near reserve. / Garder une file d'un ensemble
-    // de workers pour nourrir les slots sans consommer la reserve proche.
+    // Keep several bounded worker-sets ready so frame-paced admission can feed the
+    // background lane continuously without moving mesh application onto the game thread.
     const int32 ReadyQueueDepth = FMath::Clamp(
-        GVoxelForgeStreamingReadyQueueDepth, 0, MaxTasks);
+        GVoxelForgeStreamingReadyQueueDepth, 0, FMath::Min(64, FMath::Max(1, MaxTasks) * 4));
     const int32 TotalAdmissionCap = MaxTasks + ReadyQueueDepth;
     const int32 AuthoredNearTaskReserve = GetNearTaskReserve(MaxTasks);
     const FIntVector CenterChunk = GetTileCoordAtLocalCm(WorldToLocalCm(CenterPosition));
@@ -12787,7 +12787,7 @@ void AVoxelWorld::LoadTile(const FVoxelTileKey& Tile, EVoxelTileWorkPriority Wor
 
     const int32 MaxTasks = GetMaxConcurrentStreamingTasks();   // Core and worker lane clamped. / Core et voie worker plafonnes.
     const int32 ReadyQueueDepth = FMath::Clamp(
-        GVoxelForgeStreamingReadyQueueDepth, 0, MaxTasks);
+        GVoxelForgeStreamingReadyQueueDepth, 0, FMath::Min(64, FMath::Max(1, MaxTasks) * 4));
     const int32 TotalAdmissionCap = MaxTasks + ReadyQueueDepth;
     const bool bNearPriority = WorkPriority != EVoxelTileWorkPriority::Horizon;
     const int32 NearReserve = EffectiveNearTaskReserve >= 0
@@ -13130,7 +13130,7 @@ void AVoxelWorld::LoadCoreTile(const FVoxelTileKey& Tile, EVoxelTileWorkPriority
 
     const int32 MaxTasks = GetMaxConcurrentStreamingTasks();
     const int32 ReadyQueueDepth = FMath::Clamp(
-        GVoxelForgeStreamingReadyQueueDepth, 0, MaxTasks);
+        GVoxelForgeStreamingReadyQueueDepth, 0, FMath::Min(64, FMath::Max(1, MaxTasks) * 4));
     const int32 TotalAdmissionCap = MaxTasks + ReadyQueueDepth;
     const bool bNearPriority = WorkPriority != EVoxelTileWorkPriority::Horizon;
     const int32 NearReserve = EffectiveNearTaskReserve >= 0