[llvm] [SLP]Fix fmul/fadd fusion costs and retry FMA seeds after all blocks (PR #226117)

[Dmitry Sidorov via llvm-commits]

https://github.com/MrSidims approved this pull request.

LGTM, I believe it has to be rebased one more time to update checks in a test added in https://github.com/llvm/llvm-project/pull/226609

Side Q: I was a bit surprised to see, that SLPVectorizer/X86/redux-feed-buildvector.ll didn't require an CHECKs update with this patch. I assume it relates to ttps://github.com/llvm/llvm-project/pull/225367 where `fast` got replaced with `reassoc nnan ninf nsz arcp afn`. I expect, that with `fast` current main would still vectorize the sample and with the patch it will stop doing so. Is it desired behaviour for X86?

https://github.com/llvm/llvm-project/pull/226117