Floating point math implementation need fixes

[mzattera]

Currently, for the C64 platform, Java floating point math is disabled (as it is not supported by cc65 anyway).
There should be some work done to enable and test it, as least for platforms like Windows where float is provided.
Notice the code assuments float fits a STACKWORD, which might or might not be the case.
Probably some code should b provided for float math operation which is agnostic about how floats are implemented on the target platform.