[clang] [llvm] [AMDGPU] Implement AsyncMark Stages (PR #220442)

[Krzysztof Drewniak via cfe-commits]

krzysz00 wrote:

(and we'd probably define stage numbering to be function-local, so that we don't have to do whole-program inference on this, but still)

https://github.com/llvm/llvm-project/pull/220442