Release v0.5.2

Features

• Enhanced Delaunay triangulation wrappers.
• Updated marching tetrahedra operations.

Refactors & Build

• Refactored flood fill kernel to use memory-optimized int8_t masks and atomicCAS_int8, significantly cutting volumetric memory overhead.
• Updated mesh_bvh bindings, C++ headers, and CUDA kernels to correctly extract and cast int8_t flood fill mask pointers.
• Dynamically injected Conda CUDA target include paths in setup.py to prevent missing vector_types.h compiler errors.
• Updated project metadata and build dependencies.
• Added acknowledgements and comprehensive installation instructions to README.

Examples

• Verified int8_t flood fill tensor pipeline stability in downstream simple_contouring metrics scripts.