[llvm] [LV][AArch64] Prefer Fixed over Scalable if cost-model is equal (Neoverse V2) (PR #95819)

[David Green via llvm-commits]

davemgreen wrote:

LDP/STP was only one of the reasons to prefer fixed-width over scalable vectorization on a tie. I have been trying to think of reasons why we would want to favour scalable vectorization over fixed-width, and if we know that the scalable factor would not be higher as it will be here, then I couldn't think of anything that really make scalable vector the better choice. @david-arm I see where you are coming from with your suggestion but it feels like an over-engineered way of only solving part of the problem, and at the end of the day, when we don't know any better and can't make the decision based on anything else, fixed width should be the preferred option.

https://github.com/llvm/llvm-project/pull/95819