[llvm] [X86][AVX512] Better lowering for `_mm512_maskz_shuffle_epi32` (PR #121147)

[Simon Pilgrim via llvm-commits]

https://github.com/RKSimon commented:

Added a few comments, its tricky to decide when its worth using a constant predicate mask so we've not tried too hard to lower to them.

https://github.com/llvm/llvm-project/pull/121147