Fix _dump_float producing invalid TOML for Decimal special values

[bysiber]

Summary

_dump_float produces invalid TOML when encoding Decimal special values (Infinity, -Infinity, NaN).

Problem

_dump_float uses "{}".format(v) to convert numeric values to strings. For float, Python produces lowercase inf/nan which matches TOML's expected format. But for Decimal, Python produces Infinity/NaN (capitalized), which is not valid TOML:

>>> import toml
>>> from decimal import Decimal
>>> toml.dumps({"val": Decimal("inf")})
'val = Infinity\n'
# Invalid TOML — should be 'val = inf\n'

>>> toml.loads('val = Infinity')
# TomlDecodeError: invalid literal for int() with base 0: 'Infinity'

Since _dump_float is registered in dump_funcs for both float and Decimal, it needs to handle both formats.

Fix

Normalize Infinity → inf, -Infinity → -inf, and NaN → nan in _dump_float before applying the exponent formatting. This covers both float and Decimal inputs.