Add support for compare between float128 values.
The test case is underspecified. Please show the code that extracts the correct CR-bit from the result of the comparison. Also, please add testing for fsel as that will presumably produce select_cc in addition to all the setcc's. I presume there is additional work to do to handle the former.