FIX: reject non-finite parameters in the remaining audio and image converters

[Javier Valero (jav1er8)]

Follow-up to #2560 (AudioSpeed) and #2566 (AudioEcho). The same validation pattern is used across the converter family and four converters still have the gap: x <= 0 or x > N never rejects NaN, because every comparison against NaN is false.

Parameters that accept non-finite values on main

Converter	Parameter	NaN	inf
AudioWhiteNoiseConverter	noise_scale	accepted	rejected
AudioVolumeConverter	volume_factor	accepted	accepted
ImageColorSaturationConverter	level	accepted	accepted
ImageRotationConverter	angle	accepted	accepted

ImageRotationConverter.angle has no validation at all.

Why it matters

The output is corrupted silently, which is the bad case for a red-teaming tool: the artifact is written, stored in memory and sent to the target as if it were valid.

• AudioWhiteNoiseConverter(noise_scale=nan) — rng.normal(0, nan * max_val) makes the whole array NaN, np.clip does not repair NaN, and the int16 cast turns it into zeros. Silent audio, no error.
• AudioVolumeConverter(volume_factor=inf) — 0 * inf = nan on silent samples, everything else saturates.
• ImageColorSaturationConverter(level=nan) and ImageRotationConverter(angle=nan) — PIL returns an all-black image, no exception.

Changes

math.isfinite in each validation, plus a finiteness check on ImageRotationConverter.angle. Error messages now name finiteness, so the existing tests matching on them are updated the same way #2560 and #2566 updated theirs.

Tests follow the parametrized pattern introduced in #2560 and cover nan, inf and -inf for every affected parameter.

Verification

• pytest tests/unit/converter: 1476 passed, 34 skipped
• ruff format --check, ruff check, ty check: clean

Note on scope

This PR originally also covered AudioEchoConverter. #2566 landed that while this was open, so those two files were dropped on rebase and main's version kept — see the comment below.

[Javier Valero (jav1er8)]

@microsoft-github-policy-service agree

[Javier Valero (jav1er8)]

Rebased onto main and narrowed the scope.

#2566 landed the AudioEcho half of this while the PR was open, so delay and decay are already fixed upstream. I dropped my version of those two files entirely and kept main's — my changes there were only different wording of the same fix, and rewording error messages in an already-fixed file would be gratuitous churn.

What remains is the four converters that are still unguarded on main:

Converter	Parameter	NaN	inf
AudioWhiteNoiseConverter	noise_scale	accepted	rejected
AudioVolumeConverter	volume_factor	accepted	accepted
ImageColorSaturationConverter	level	accepted	accepted
ImageRotationConverter	angle	accepted	accepted

ImageRotationConverter.angle still has no validation at all.

Now +57/-21 across 8 files instead of +86/-33 across 10. pytest tests/unit/converter is green (1476 passed, 34 skipped), ruff format --check and ruff check clean.