Count every boolean category in CategoryCount

[breken-ai]

What was wrong

CategoryCountCalculation._calculate_value has a special branch for when every category is a boolean:

if all(isinstance(c, bool) for c in self.metric.categories):
    #  only one boolean label is possible here
    value = counts[self.metric.categories[0]]

CategoryCount accepts several categories, so categories=[True, False] is valid, but that branch counts only the first one:

df = pd.DataFrame({"b": [True, True, False, False]})
Report([CategoryCount(column="b", categories=[True, False])]).run(Dataset.from_pandas(df))
# before: {'count': 2.0, 'share': 0.5}
# after:  {'count': 4.0, 'share': 1.0}

Fix

Remove the boolean branch. The loop below it already looks up each label with cat in counts / counts[cat], which works for boolean labels too. A boolean label on a 0/1 integer column still counts 0, as before.

Tests

• New tests/future/metrics/test_category_count.py covers multiple string categories (including a missing one) and boolean categories. [True, False] and [False, True] with a None fail on main (count 2 instead of 4, and 1 instead of 3). All 6 cases pass with the fix.
• The broader suite (tests/calculations tests/future tests/metrics tests/metric_preset tests/test_preset tests/test_suite tests/stattests) has the same 71 pre-existing failures as main (all No module named 'IPython', because I did not install the jupyter extra), and no new ones.
• ruff check and ruff format --check pass on the changed files.

Side note: the existing tests/future/metrics/category_count.py is not collected by pytest because of its file name, and it fails on main (it compares CountValue.count to an int). I left it unchanged.

This PR was written with an AI coding agent (Claude Code, run by breken-ai), and I checked the reproduction and the tests above.