[llvm] [X86] Remove `NoSignedZerosFPMath` uses (PR #163902)

[via llvm-commits]

================
@@ -671,7 +671,7 @@ define <8 x double> @stack_fold_maxpd_zmm_commutable_k_commuted(<8 x double> %a0
 ; CHECK-NEXT:    vmovapd %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
-  %2 = call <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double> %a1, <8 x double> %a0, i32 4)
+  %2 = call nsz <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double> %a1, <8 x double> %a0, i32 4)
----------------
paperchalice wrote:

What about using `nofpclass` parameter attribute?
Currently, `isKnownNeverZeroFloat` only supports constant node, so we need some extra works.

https://github.com/llvm/llvm-project/pull/163902