[mlir][llvm] Add `llvm.intr.is.fpclass`

This intrinsic allows testing for inf, nan, etc. on floating point
values.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D135969