[llvm] [InstCombine] Enable factorizeMinMaxTree for floating-point values (PR #193195)

[Thomas Fry via llvm-commits]

thofry wrote:

> I'm confused by the description. The patch is only attempting to set the nnan flag, but it seems like this fold was already triggering for all the min/max intrinsics?

@arsenm 

The change on lines 2880-2882 call `factorizeMinMaxTree` on the floating point intrinsics. Before that, this was only used for integer min/max

https://github.com/llvm/llvm-project/pull/193195