Keep whole numbers finite when they do not fit in a float

[ShamikOfficial]

Fixes #419

float() was used to test for NaN and infinity. For an integer larger than the float range that raises OverflowError. For the same number written as a digit string, float() returns infinity, so the value was reported as +Inf.

Whole numbers are now treated as finite. Scientific notation for integers that do not fit in a float is formatted with Decimal, so it does not overflow.

Tested with pytest tests/test_number.py (242 passed).

[ShamikOfficial]

I don't have permission to add labels. Please add changelog: Fixed when you have a moment so the changelog check can pass.

[Wkhanex]

The huge integer and decimal-string controls pass, but this head defines _nonfinite_float twice with identical bodies. I confirmed two top-level definitions in the AST of src/humanize/number.py; Python silently overwrites the first definition with the second. Please remove the duplicate before merging so there is one implementation to maintain and no redefinition lint finding. I also executed scientific/apnumber controls for 10**400 and its decimal string; both remained finite and exact. AI-assisted review on Python 3.12 with English locale hooks, not the full package suite.