Update CudaArch structs and functions in clang to something more generic

[npmiller]

The CudaArch struct and functions include AMD GPU values in:

• https://github.com/intel/llvm/blob/sycl/clang/lib/Basic/Cuda.cpp#L73

It would be good to revisit this and change it to something more generic like maybe GPUArch or something like that as the Cuda tag doesn't really reflect what it is anymore.

This was discussed on the following PR:

• [SYCL][ROCm] Use offload-arch instead of mcpu for AMD arch #4239 (comment)

[jchlanda]

Upstream PR: llvm/llvm-project#97028