[clang] [Clang] Introduce scoped variants of GNU atomic functions (PR #72280)

[Joseph Huber via cfe-commits]

jhuber6 wrote:

> Just a FYI, that recent NVIDIA GPUs have introduced a concept of [thread block cluster](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#thread-block-clusters). We may need another level of granularity between the block and device.

Should be easy enough, though the numbers would no longer be incremental if we put it between there. It's somewhat difficult to decide what these things should be called. Also I was somewhat tempted to keep the names all the same length like the `__ATOMIC` ones are, but that might not be worth the effort.

That being said, As far as I'm aware the Nvidia backend doesn't handle scoped atomics at all yet, we simply emit `volatile` versions even when scopes exist in PTX.

https://github.com/llvm/llvm-project/pull/72280