quad-float128.h (IBM128_TYPE): Explicitly use __ibm128, instead of trying to use...