fix(validation): correct 'and' to 'or' in argument validation across multiple metrics

[AbdullahRasheed45]

Summary

Several argument validation guards across multiple metrics used the wrong logical operator ( instead of ). As a result, invalid arguments were silently accepted and caused cryptic runtime errors deep inside the metric computation rather than a clear ValueError at initialisation time.

This pattern was first identified in stat_scores.py (issues #3405, resolved in PRs #3406/#3416) and in retrieval metrics (issue #3378). This PR fixes the identical bug in all remaining affected files.

The Bug

The guards follow the pattern:

# Wrong: only fails when BOTH conditions are True
if not isinstance(x, SomeType) and x_is_invalid:
    raise ValueError(...)

For an invalid-value-but-correct-type input (e.g., min_sensitivity=2.0):

• not isinstance(2.0, float) → False
• 2.0 is outside [0, 1] → True
• False and True → False — no error raised!

Fix: replace with so validation fails when either the type or the value is wrong.

Files Changed

File	Parameter	Invalid values that bypassed validation
functional/classification/specificity_sensitivity.py (3 sites)	min_sensitivity	out-of-range floats, e.g. 2.0, -0.5
functional/classification/recall_fixed_precision.py (3 sites)	min_precision	out-of-range floats, e.g. 2.0, -0.1
functional/classification/sensitivity_specificity.py (3 sites)	min_specificity	out-of-range floats, e.g. 2.0, -0.5
regression/spearman.py	num_outputs	0, negative integers
regression/pearson.py	num_outputs	0, negative integers
regression/log_cosh.py	num_outputs	0, negative integers
classification/group_fairness.py (2 sites)	num_groups	0, 1, negative integers
image/psnrb.py	block_size	0, negative integers

Minimal Reproduction

from torchmetrics.functional.classification import binary_specificity_at_sensitivity

# Before fix: no error raised for out-of-range float
binary_specificity_at_sensitivity(preds, target, min_sensitivity=2.0)

# After fix: raises ValueError immediately
# ValueError: Expected argument `min_sensitivity` to be an float in the [0,1] range, but got 2.0

from torchmetrics.regression import SpearmanCorrCoef

# Before fix: no error for zero num_outputs (int)
SpearmanCorrCoef(num_outputs=0)

# After fix: raises ValueError immediately
# ValueError: Expected argument `num_outputs` to be an int larger than 0, but got 0