Fix IEEE 754 comparison ((>) and (>=)) for BigFloat when NaN is involved

[lsrcz]

This pull request fixes the (>) and (>=) operator of the BigFloat, which aren't IEEE 754 complaint when NaN is involved.

The original implementation relies on the default implementation of the (>) and (>=). Unfortunately, the base library implemented them as negation of (<) and (<=), which would make nan > 1 or 1 > nan evaluate to True.

[RyanGlScott]

Thanks, @lsrcz! Would you be willing to add a test case here that demonstrates that the fix works?

[lsrcz]

Sure, let me write some test cases soon.

[lsrcz]

Hi @RyanGlScott , I've added some tests for the comparison operators with NaN.

[RyanGlScott]

I've uploaded libBF-0.6.8 to Hackage with these changes.

[lsrcz]

Thank you for all the quick response!