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

[Krzysztof Drewniak via cfe-commits]

krzysz00 wrote:

And in that example, the second asyncmark captures *both* the async load to lds and the TDM instruction - that's how it's supposed to work

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