[docs] Mention inline CUDA extensions in tutorial

[vadimkantorov]

It would be nice to mention load_inline (pytorch/pytorch#7059) in the tutorial
https://pytorch.org/tutorials/advanced/cpp_extension.html

A practical use-case for an example may be a bilateral filter from https://github.com/NVIDIA/FastPhotoStyle/blob/master/smooth_filter.py (currentlly uses CuPY and pynvrtc).

cc @jlin27 @mruberry

[vadimkantorov]

At least load_inline now appears in docs

[vadimkantorov]

Although, caching of compiled binaries was recently still broken: pytorch/pytorch#68066