[te][llvm] Enable fused multiply-add (fma) in code generation (#45906)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/45906
Test Plan: Imported from OSS
Reviewed By: SplitInfinity
Differential Revision: D24142404
Pulled By: bertmaher
fbshipit-source-id: a8db2e66c1e65bbb255886e165a1773723cbcd20