Creates a double-precision floating-point number from a higher order word and a lower order word.