fix(intcomma): handle scientific notation for large/small floats

[Jah-yee]

Summary

Fixes #431 — intcomma() returns scientific notation for floats of 1e16 and above (and very small floats like 1e-5), with no commas added.

Root cause: Python's repr() switches to scientific notation for floats with magnitude >= 1e15 or <= 1e-5. The f"{value:,}" format specifier uses repr() internally, so the comma formatting is never applied to the exponent form.

Fix: When ndigits is None and the float is a whole number, format as int to avoid trailing .0. When the float would be formatted in scientific notation, use Decimal for a proper decimal representation before applying commas.

Before:

>>> intcomma(1e16)
'1e+16'
>>> intcomma(1.5e20)
'1.5e+20'

After:

>>> intcomma(1e16)
'10,000,000,000,000,000'
>>> intcomma(1.5e20)
'150,000,000,000,000,000,000'

Files changed:

• src/humanize/number.py: Fixed intcomma() to handle scientific notation in float repr

[rmz-oz]

this drops the .0 from every whole float: intcomma(1.0) is now '1' and 1234.0 gives '1,234', but #431 expects '10,000,000,000,000,000.0'. fwiw #432 keeps both the .0 and the sign by going through f"{Decimal(str(value)):,f}". no tests for the new branches either.

[rmz-oz]

repr already switches to an exponent below 1e-4, so with this bound intcomma(5e-5) and intcomma(1.5e-5) still come back as '5e-05' and '1.5e-05'.

[rmz-oz]

int("-0") is 0, so small negatives lose the sign here: intcomma(-1e-5) gives '0.00001' and intcomma(-2.5e-7) gives '0.00000025'.