CLab

perf(culling): optimize hardware occlusion culling implementation

• Increased initial query pool size from 512 to 1024 for better performance
• Added shader caching mechanism to reduce redundant shader operations
• Implemented matrix caching for projection and model-view transformations
• Replaced direct shader drawing with optimized draw calls
• Removed redundant imports and cleaned up code structure
• Moved block entity culling logic to separate mixin class
• Added proper shader matrix updates during query batches
• Implemented resource cleanup in endQueryBatch methods