FURB157 has false negatives for non-finite float strings

[dscorbett]

verbose-decimal-constructor (FURB157) has false negatives in Ruff 0.8.0 for some non-finite float strings. It recognizes the following strings case-insensitively in expressions like Decimal(float("inf")), but no others:

https://github.com/astral-sh/ruff/blob/a90e404c3f010446ab8c18b4793c78834eeb65b7/crates/ruff_linter/src/rules/refurb/rules/verbose_decimal_constructor.rs#L157

float strips white space. For example, Decimal(float(" nan ")) should be simplified to Decimal(" nan "). It could be further normalized to Decimal("NaN"), but FURB157 does not currently normalize other float strings, so I think it should keep the string argument unchanged.

float allows both signs for all non-finite values. Decimal(float("+inf")), Decimal(float("+infinity")), Decimal(float("+nan")), and Decimal(float("-nan")) should be simplified. FURB157 should not normalize these string arguments either, except for the last one. Decimal(float("-nan")) is equivalent to Decimal("nan"), not Decimal("-nan"), so for an argument of "-nan", the sign character should be deleted.

[dylwil3]

Decimals strike again! Thanks for finding this, we'll get it fixed 😄