Add configurable bounds for hash-grid input dimensions

[ayushozha]

Summary

• add configurable TCNN_HASH_MIN_DIM and TCNN_HASH_MAX_DIM build options for hash grids
• thread the same dimension bounds through CMake and the PyTorch extension build
• add test coverage and README usage notes for enabling 5D hash grids

Why

Issue #422 asks for 5D grid support. The current hash-grid dispatch is hard-coded to a narrow range of input dimensions, so enabling 5D grids requires rebuilding the library. This change makes the supported hash-grid dimension range explicit and configurable at build time while preserving the existing defaults.

Validation

• python -m py_compile bindings/torch/setup.py
• cmake -S . -B build-hash-dim-check -G Ninja -DTCNN_BUILD_TESTS=1 -DTCNN_HASH_MAX_DIM=5
• cmake --build build-hash-dim-check --target test_grid
• ctest --output-on-failure -V -R test_grid

Closes #422.

[Tom94]

Thanks for the PR, however I'm not a fan of making this configurable in the first place. It's a lot of boilerplate for something the user could enable with a very simple patch file / fork of tcnn.