[llvm] [SDAG] Heed enableAggressiveFMAFusion when folding fma(a,b,c*d)+e (PR #94209)

[Jay Foad via llvm-commits]

jayfoad wrote:

> IIRC D28332 was to help avoid fp bottlenecks - as many x86 cpus process fadd + fmul/fma on separate fp pipes.

OK, but wouldn't the obvious solution be to turn off enableAggressiveFMAFusion if you don't want it?

https://github.com/llvm/llvm-project/pull/94209