[llvm] [X86] Fold EXPAND(X,Y,M) -> SELECT(M,X,Y) when M is a lowest bit mask (PR #179630)

[Simon Pilgrim via llvm-commits]

================

----------------
RKSimon wrote:

Yes - lower1BitShuffle is missing basic AND blend masking support that would avoid all the extensions etc. and should probably eventually simplify to movmsk

https://github.com/llvm/llvm-project/pull/179630