[llvm] [InstCombine] Transform high latency, dependent FSQRT/FDIV into FMUL (PR #87474)

[Sushant Gokhale via llvm-commits]

sushgokh wrote:

> I think this looks good now. Thanks for your perseverance.

Thanks @andykaylor for diligently looking into this and helping me bring it to the approval stage :) 

https://github.com/llvm/llvm-project/pull/87474