Use ensure_all_finite from scikit-learn 1.6.0 to avoid FutureWarning

[yhay81]

Fixes #41.

check_X_y_wrapper / check_array_wrapper switched from force_all_finite to ensure_all_finite only at scikit-learn >= 1.8.0, but scikit-learn introduced ensure_all_finite in 1.6.0 and keeps force_all_finite as a deprecated alias (with a FutureWarning) until 1.8. This changes the threshold to 1.6.0 so that scikit-learn 1.6.x / 1.7.x no longer print the warning on every fit().

Verified on scikit-learn 1.6/1.7 (Kaggle image, TabM_D_Classifier.fit): the warning disappears and results are unchanged (the argument value is passed through unmodified).

🤖 Generated with Claude Code

[dholzmueller]

Thanks! Some tests are failing for unrelated reasons:

• Python 3.9 ones for some hatch compatibility issue (I would just drop python 3.9 support from pyproject.toml and the .github/workflows/testing.yml)
• macos tests for some OpenMP-related issue. ChatGPT suggests as a fix to add to .github/workflows/testing.yml an OpenMP install like this:

- uses: actions/setup-python@v5
  with:
    python-version: ${{ matrix.python-version }}

- name: Install libomp
  if: runner.os == 'macOS'
  run: brew install libomp

- name: Install hatch
  run: uv pip install --system hatch

- name: Run tests
  run: hatch test