Validate each element in ClassLabel.int2str

[shashvat-singham]

Partially fixes #8481

Problem

int2str validates the container it is given, but never the elements inside it. Three consequences, in decreasing order of severity:

>>> cl = ClassLabel(names=["neg", "pos"])

>>> cl.int2str([1.7])
['pos']                    # silently truncated by the int(v) used to index _int2str

>>> cl.int2str(["1"])
TypeError: '<=' not supported between instances of 'int' and 'str'

>>> cl.int2str([99.5])
ValueError: Unknown format code 'd' for object of type 'float'

The first is the one worth fixing: it returns a plausible-looking label for a value that was never a valid class id, with no error. The third is the existing f"...{v:d}" raise failing on the very value it was trying to report.

The TypeError case is the nested form of what #8416 fixes at the top level — that PR guards int2str("1"), this one covers int2str(["1"]).

Change

Validate each element in the loop that already walks them. A value is accepted if it converts to an integer exactly, so this keeps working for every input the docstring promises:

cl.int2str([0, 1])                        # ['neg', 'pos']
cl.int2str(np.array([0, 1]))              # ['neg', 'pos']
cl.int2str([np.int64(0), np.int64(1)])    # ['neg', 'pos']

I deliberately checked integrality (int(v) != v) rather than isinstance(v, int), because iterating a numpy array or a torch/tf tensor yields scalar objects that are not Python ints — a type check would have broken the framework-tensor inputs the docstring explicitly supports.

Scope note

#8481 also reports str2int([1]) returning [1]. Having read the code I think that one is more defensible than I made it sound in the issue: _strval2int does str(value) first and numeric strings like "1" are deliberately supported, so an integer falling through is a natural consequence of that feature rather than a separate bug. I've left it alone — happy to revisit if you read it differently.

Tests

test_classlabel_int2str_validates_elements covers the valid inputs (list, numpy array, numpy scalars) and each rejected case. It fails on main and passes with the change.

$ pytest tests/features/test_features.py -q
201 passed, 3 skipped

ruff format --check and ruff check are clean on both files.