Use LLVM `isnan` intrinsic to implement SLEIGH's `FLOAT_NAN`

[tetsuo-cpp]

As of #585, we're implementing FLOAT_NAN with our own code that checks the float against itself with an ordered float comparison.

LLVM is considering an isnan intrinsic that would be helpful for us. It got merged and then subsequently reverted in https://github.com/llvm/llvm-project/commit/3f1f08f0ed6aa1ebd4678583cbbf5026ca5cbdf8 but it'd be good to keep an eye on this and see whether it makes its way into LLVM 15.