Fix inverted is_all_nan / is_any_nan / is_all_infinite / is_any_infinite

[youdie006]

Fixes #1612.

Root cause

The boolean_ops! macro in src/numeric/impl_float_maths.rs had an inverted $all predicate: Zip::from(self).all(|&elt| !elt.$func()) tests "all elements are NOT NaN", the opposite of the documented "true if all elements are NaN". $any was defined as !self.$all(), so the double negation made is_any_* appear correct while is_all_* returned true for an all-finite array.

Fix

Both predicates are corrected together (fixing only $all would break the is_any_* methods that currently ride on the double negation):

• $all -> Zip::from(self).all(|&elt| elt.$func())
• $any -> Zip::from(self).any(|&elt| elt.$func())

This fixes is_all_nan, is_any_nan, is_all_infinite, and is_any_infinite in one change. Empty-array semantics are preserved (all -> true, any -> false).

@dbranford correctly identified the $all flip and raised the is_any question in the issue thread; this PR applies both corrections and adds regression coverage.

Tests

New tests/nan_all_repro.rs covers both the NaN and infinite families with f64: all-finite (is_all_* and is_any_* false), all-nan/all-inf (is_all_* and is_any_* true), and mixed (is_all_* false, is_any_* true).

Red-green verified: with the fix reverted, the first assertion (!all_finite.is_all_nan()) panics; with the fix, both tests pass. cargo fmt (nightly, per repo rustfmt.toml), cargo clippy (no new warnings), and the existing tests/numeric.rs suite all pass.

---

AI-assisted: this change was prepared with the help of an AI coding assistant and reviewed by me before submission.

[akern40]

@youdie006 thank you for your contribution! Sorry for the delay, life got busy but I'm trying to get back to the repo now