[llvm] [LSV][AMDGPU] Vectorize unordered and monotonic atomic loads (PR #190152)

[Drew Kersnar via llvm-commits]

dakersnar wrote:

FYI we have been working on an LSV change that removes scalar type size as an equivalence class parameter: https://github.com/llvm/llvm-project/pull/177908, allowing for mixed type size vectorization. This probably doesn't conceptually conflict with your change, but wanted to call it out in case it does. cc @gandhi56 @cmc-rep for visibility.

https://github.com/llvm/llvm-project/pull/190152