Set RPATH for cuda libraries from python package

[take-cheeze]

Description

Fixes crashs for binary linked with both libtorch and libtransformer_engine running with nsys profile .
It was caused by wrong libcudnn.so loaded when system package like libcudnn9-cuda-12 is installed current Transfomer Engine use it instead of nvidia-cudnn-cu12 pypi package.
Though by setting export LD_LIBRARY_PATH=$(pip show nvidia-cudnn-cu12 | awk '/Location:/ { print $2 }')/nvidia/cudnn/lib maybe a workaround.

May also need NVIDIA/cudnn-frontend#180

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

• Finds Python before including cuDNN.cmake from cudnn-frontend
• Sets CMAKE_INSTALL_RPATH like pytorch does: https://github.com/pytorch/pytorch/blob/e2c683458476be48569098d74cb9fd41315376e9/.ci/manywheel/build_cuda.sh#L248-L260

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes