[X86] Failure to handle bf16 strict_fsqrt

[RKSimon]

define <8 x bfloat> @strict_fsqrt(<8 x bfloat> %a) strictfp {
  %sqrt = call <8 x bfloat> @llvm.experimental.constrained.sqrt.v8bf16(
                              <8 x bfloat> %a,
                              metadata !"round.dynamic",
                              metadata !"fpexcept.strict") #0
  ret <8 x bfloat > %sqrt
}

Noticed while working on tests for #165682 / #167692

[llvmbot]

@llvm/issue-subscribers-backend-x86

Author: Simon Pilgrim (RKSimon)

``` define <8 x bfloat> @strict_fsqrt(<8 x bfloat> %a) strictfp { %sqrt = call <8 x bfloat> @llvm.experimental.constrained.sqrt.v8bf16( <8 x bfloat> %a, metadata !"round.dynamic", metadata !"fpexcept.strict") #0 ret <8 x bfloat > %sqrt } ```

Noticed while working on tests for #165682 / #167692

[phoebewang]

This duplicates #164152. We cannot support bf16 strict FP operations due to all the bf16 instructions don't raise exception and ignore rounding control. We should diagnosis and stop generate them in the front end.