Fix intcomma() returning scientific notation for large floats

[Vinayak19112003]

Closes #431.

intcomma() used f"{value:,}" for floats, which formats via repr(). That switches to scientific notation at 1e16 (and for tiny floats like 1e-5), so no thousands separators were added:

>>> intcomma(1e16)
'1e+16'

The float is now expanded to fixed-point via Decimal(repr(value)) before grouping, so the full decimal representation gets commas:

>>> intcomma(1e16)
'10,000,000,000,000,000.0'
>>> intcomma(1.5e20)
'150,000,000,000,000,000,000.0'
>>> intcomma(0.00001)
'0.00001'

The ndigits path already used fixed-point formatting and is unchanged, as is the integer path.

Tests: new test_intcomma_large_float_no_scientific_notation (6 cases: 1e15/1e16/1.5e20/negative/tiny/ndigits=0) — fails on main, passes with the fix. Full tests/test_number.py: 247 passed. Ruff check and format clean.

[Wkhanex]

Additional value-preservation check: six large, small, signed and signed-zero floats (1e16, -1e16, 1.5e20, 1e-5, -1e-5, -0.0) all rendered without scientific notation. Parsing the output after removing grouping commas reproduced Decimal(repr(value)) exactly. This checks that expanding the shortest float representation adds grouping without introducing extra rounding, including negative tiny values. I ran these checks on Python 3.12 with English locale hooks; no full localization or pytest run.