[llvm] [InstCombine] Canonicalize nsz NaN-fallback float clamp to minnum/maxnum (PR #213690)

[Jun Yeong Kim via llvm-commits]

junyeong0619 wrote:

Fixed the crash: restricted the fold to isFPOrFPVectorTy() so FP aggregate selects (e.g. { float, float }) are excluded. Added n11_fp_aggregate as a regression test.

https://github.com/llvm/llvm-project/pull/213690