manual: clarify defintions of floating point exponent bounds (bug 31518)