[llvm] [AMDGPU] Add basic patterns to select lshl_or instead of v_perm (PR #65693)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm commented:

Can we avoid emitting a perm in the first place? Can we decompose existing perms into the shift + or? Does computeKnownBits know about AMDGPUperm?

https://github.com/llvm/llvm-project/pull/65693