[llvm] [LV][AArch64]: Utilise SVE ld4/st4 instructions via auto-vectorisation (PR #89018)

[Eli Friedman via llvm-commits]

efriedma-quic wrote:

Given the way the pass pipeline is structured, cost modeling in the vectorizer itself tends to be more important than modeling in subsequent passes.  I guess maybe it's not a big deal what the vectorizer generates if the vectorizer itself has some way to get the correct numbers.

https://github.com/llvm/llvm-project/pull/89018