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

[Igor Kirillov via llvm-commits]

igogo-x86 wrote:

@DavidSpickett, yes, I reverted the commit. mecmp(..., ..., 3) on certain data was not correct. Going to investigate how to handle it properly

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