ARM & PYTORCH 2.6 &CUDA 12.8

[johnnynunez]

This pull request includes updates to CUDA dispatch macros and improvements to the setup.py script for the Kaolin library. The most important changes include updating the dispatch macros to use scalar_type() and enhancing the setup.py script for better version handling and logging.

CUDA dispatch macro updates:

• Updated AT_DISPATCH_FLOATING_TYPES_AND_HALF macros to use scalar_type() instead of type() in kaolin/csrc/ops/spc/point_utils_cuda.cu (interpolate_trilinear_cuda_impl, coords_to_trilinear_cuda_impl) [1] [2].
• Updated AT_DISPATCH_FLOATING_TYPES_AND_HALF macros to use scalar_type() instead of type() in kaolin/csrc/ops/spc/query_cuda.cu (query_cuda_impl, query_multiscale_cuda_impl, query_cuda_impl_empty) [1] [2] [3].
• Updated AT_DISPATCH_FLOATING_TYPES_AND_HALF and AT_DISPATCH_INTEGRAL_TYPES macros to use scalar_type() in kaolin/csrc/render/spc/raytrace_cuda.cu (mark_pack_boundaries_cuda_impl, diff_cuda_impl, sum_reduce_cuda_impl, cumsum_cuda_impl, cumprod_cuda_impl) [1] [2] [3] [4] [5] [6] [7].

Setup script improvements:

• Enhanced setup.py to better handle version constraints and logging for PyTorch and Cython dependencies [1] [2].
• Improved the handling of CUDA architectures and cross-compilation settings in setup.py [1] [2].

[johnnynunez]

It is working fine on jetson agx orin cuda 12.8 pytorch 2.6.0

https://github.com/dusty-nv/jetson-containers/tree/master/packages/pytorch/torch3d