Incremental terrain upload + M2 instance dedup hash for city stutter

Terrain finalization was uploading all 256 chunks (GPU fence waits) in one
atomic advanceFinalization call that couldn't be interrupted by the 5ms time
budget. Now split into incremental batches of 16 chunks per call, allowing
the time budget to yield between batches.

M2 instance creation had O(N) dedup scans iterating ALL instances to check
for duplicates. In cities with 5000+ doodads, this caused O(N²) total work
during tile loading. Replaced with hash-based DedupKey map for O(1) lookups.

Changes:
- TerrainRenderer::loadTerrainIncremental: uploads N chunks per call
- FinalizingTile tracks terrainChunkNext for cross-frame progress
- TERRAIN phase yields after preload and after each chunk batch
- M2Renderer::DedupKey hash map replaces linear scan in createInstance
  and createInstanceWithMatrix
- Dedup map maintained through rebuildSpatialIndex and clear paths

include/rendering/terrain_renderer.hpp

@@ -86,6 +86,13 @@ public:
                      const std::vector<std::string>& texturePaths,
                      int tileX = -1, int tileY = -1);
 
+    /// Upload a batch of terrain chunks incrementally. Returns true when all chunks done.
+    /// chunkIndex is updated to the next chunk to process (0-255 row-major).
+    bool loadTerrainIncremental(const pipeline::TerrainMesh& mesh,
+                                const std::vector<std::string>& texturePaths,
+                                int tileX, int tileY,
+                                int& chunkIndex, int maxChunksPerCall = 16);
+
     void removeTile(int tileX, int tileY);
 
     void uploadPreloadedTextures(const std::unordered_map<std::string, pipeline::BLPImage>& textures);