[llvm] [AMDGPU] Fix simplification of VM_VSRC based on VMEM waits (PR #178711)

[Ryan Mitchell via llvm-commits]

https://github.com/RyanRio approved this pull request.

I think the last time I didn't have review permissions haha.

Anyhow, 2 nit/concern, feel free to address as you see fit, approach still looks correct to me.

https://github.com/llvm/llvm-project/pull/178711