[llvm] [AArch64] Add MATCH loops to LoopIdiomVectorizePass (PR #101976)

[Sjoerd Meijer via llvm-commits]

sjoerdmeijer wrote:

Thanks for the experiments and data, I think that looks really good! This always looks like a win, except when the size is or 1 or 2. That is most likely so exceptional we don't have to consider runtime size checks to either fall into the scalar or the vector loop.  

https://github.com/llvm/llvm-project/pull/101976