[POWERPC] Fixup error handling when emulating a floating point instruction