add __cuda_array_interface__

[casperdcl]

• v3: https://numba.readthedocs.io/en/latest/cuda/cuda_array_interface.html
• v2: https://numba.pydata.org/numba-doc/dev/cuda/cuda_array_interface.html
  • e.g.: Implement Tensor.__cuda_array_interface__. pytorch/pytorch#11984
  • e.g.: Implement __cuda_array_interface__ cupy/cupy#1144

This would allow easily adding to https://github.com/xnd-project/arrayviews (#23):

• CPU arrays (already possible)
• CUDA arrays, thereby supporting e.g.: cupy, numba, pyarrow, xnd, cudf, ...

Other references:

• https://github.com/rapidsai