nir: Temporarily disable optimizations for MSVC ARM64 in constant folding
There's currently an MSVC optimizer bug which causes a stack overflow in the compiler if it attempts to optimize the constant folding for fsat.
This is currently the only patch we're carrying downstream to be able to ship our ARM64 binaries, so it'd be nice to get it upstream, even though it's pretty ugly. We'll obviously remove it once the compiler bug is fixed.