fix: lower python 3.14 scikit-learn floor

[stefan-jansen]

Summary

• lower the Python 3.14 floor from to
• keep the 3.14 compiled stack installable while avoiding unnecessary downstream conflicts
• fix pandas-3.0 dtype handling in

Verification

• uv run ruff check --no-fix src tests .github
• uv run ruff format --check src tests .github
• uv run ty check
• uv run --python 3.14 pytest tests/relationships/test_correlation.py::TestFeatureSelection::test_non_numeric_feature_error -q
• exact-minimum check clone with on Python 3.14: before the repo's known Numba teardown abort ()

[Copilot]

Pull request overview

This PR improves Python 3.14 installability by lowering the scikit-learn minimum version for Python 3.14 and adjusts correlation feature type checking to be compatible with pandas 3.0 dtype behavior.

Changes:

• Lowered the Python 3.14 scikit-learn minimum from >=1.8.0 to >=1.7.2 in both pyproject.toml and uv.lock.
• Updated correlation feature validation to use pandas.api.types.is_numeric_dtype instead of numpy.issubdtype for more robust dtype handling (notably for pandas extension dtypes).

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated no comments.

File	Description
uv.lock	Aligns the locked dependency metadata with the lowered Python 3.14 scikit-learn floor.
src/ml4t/engineer/relationships/correlation.py	Fixes numeric dtype detection to avoid pandas 3.0 dtype/extension dtype issues during feature validation.
pyproject.toml	Lowers the declared Python 3.14 scikit-learn minimum requirement to reduce downstream conflicts.
CHANGELOG.md	Documents the Python 3.14 dependency floor adjustment in the release notes.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.