Change MAKE-FLOAT to multiply by the sign AFTER we've created the

quad-double.  This way we create signed quad-double zeroes correctly.
Previously, the sign was applied to the rational which loses the sign
if the rational is 0.