[llvm] [AArch64] Make use of byte FPR stores for bytes extracted from vectors (PR #134117)

[Benjamin Maxwell via llvm-commits]

MacDue wrote:

It's not quite simplified (as the fold is a little more complex in general), but I've been able to rewrite this (and the previous fold) into a unified fold that works the same for all types. This has added quite a bit more test churn though, though it mostly seems innocuous.

I've split out one patch to regenerate the `st1` tests (as previously they were manually written tests, which are annoying to update). The ~same checks can be generated using the `--filter` option of update_test_checks, patch here: https://github.com/llvm/llvm-project/pull/134919#issuecomment-2787494830 

https://github.com/llvm/llvm-project/pull/134117