Handling nan and inf values

[rocky]

a = 1e300 * 1e300 * 0
b = -1e300 * 1e300 * 0
c = 1e300 * 1e300
d = -1e300 * 1e300

gives:

a = nan
b = nan
c = inf
d = -inf

but Python doesn't allow these in source code to represent their internal values.

uncompyle6 uses float('nan'), etc. although there are numerous other ways to handle.