[llvm] [AMDGPU] Enable sinking of free vector ops that will be folded into their uses (PR #162580)

[Gheorghe-Teodor Bercea via llvm-commits]

doru1004 wrote:

@arsenm this PR has been rebased on the pre-committed test so now you can see the op sink very clearly. Awaiting your review.

https://github.com/llvm/llvm-project/pull/162580