[llvm] [CodeGen] Avoid sinking vector comparisons during CodeGenPrepare (PR #113158)

[David Green via llvm-commits]

https://github.com/davemgreen commented:

Sorry for the delay I was going through things in a strange order. I had previously only checked the test on AArch64. Do we need to worry about the extra xtn/sshll this introduces? I was going to suggest we just remove them post-isel, but that might require getting the top bits correct which requires a lot of infrastructure that currently doesn't exist.

https://github.com/llvm/llvm-project/pull/113158