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

[Alexey Bataev via llvm-commits]

alexey-bataev wrote:

> [public-lbm-reproducer.tar.gz](https://github.com/user-attachments/files/32861306/public-lbm-reproducer.tar.gz)
> 
> Here is a reproducer for the slowdown across these PRs: [PR #226268](https://github.com/llvm/llvm-project/pull/226268) [PR #226117](https://github.com/llvm/llvm-project/pull/226117) [PR #227083](https://github.com/llvm/llvm-project/pull/227083) Reverting those three patches, I see ~10% speedup on Graviton4.

Will fix ASAP

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