[llvm] [CodeGen] Improve ExpandMemCmp for more efficient non-register aligned sizes handling (PR #69942)

[David Spickett via llvm-commits]

DavidSpickett wrote:

Linaro is seeing a test suite failure on this change: https://lab.llvm.org/buildbot/#/builders/197/builds/10630

This bot is AArch64 SVE vector length agnostic, however, the change hasn't hit the other bots yet. If we're lucky it's a general issue and not SVE specific.

https://github.com/llvm/llvm-project/pull/69942