[Feature] XGBoost native multiquantile regression

[ozink-u8]

Checklist before merging this PR:

• Mentioned all issues that this PR fixes or addresses.
• Summarized the updates of this PR under Summary.
• Added an entry under Unreleased in the Changelog.

Fixes #3047

Summary

Analogously to #3032 adds multiquantile regression to XGBoost.

Other Information

Change Summary (Claude Code)

• darts/models/forecasting/xgboost.py: accept likelihood="multiquantile", set objective="reg:quantileerror" and quantile_alpha=<list>, update docstrings
• darts/utils/likelihood_models/sklearn.py: normalize quantiles to float to avoid XGBoostError from np.float64 serialization. It seemed like XGBoost had some for of casting but the sort() would create a list[np.float64] instead of a np.ndarray so their check for a np.ndarray would not catch it.
• darts/tests/models/forecasting/test_sklearn_models.py: add test_model_construction_multiquantile_xgb and test_get_estimator_multiquantile_xgb analogously to [Feature] CatBoost MultiQuantile support #3032
• examples/20-SKLearnModel-examples.ipynb: mention XGBModel alongside CatBoostModel in the multiquantile tip analogously to [Feature] CatBoost MultiQuantile support #3032
• darts/tests/explainability/test_shap_explainer.py: fix deprecated pandas freq "d" → "D" (coincidentally also analogously to [Feature] CatBoost MultiQuantile support #3032)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.72%. Comparing base (9c62525) to head (a9cf234).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3056      +/-   ##
==========================================
- Coverage   95.78%   95.72%   -0.07%     
==========================================
  Files         158      158              
  Lines       17290    17293       +3     
==========================================
- Hits        16562    16554       -8     
- Misses        728      739      +11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dennisbader]

Thanks a lot @ozink-u8, it looks great already :)

I added some minor comments, after that it's ready to be merged 🚀

[dennisbader]

This test is pretty much the same as the one for catboost above. In general we try to avoid code duplication as much as we can :)

We coud cover both models in the existing test.

You can achieve this adding another level of parametrization to test function decorators:

    @pytest.mark.skipif(
        not XGB_AVAILABLE and not CB_AVAILABLE,
        reason="XGBoost or CatBoost required for this test"
    )
    @pytest.mark.parametrize(
        "model_cls,model_kwargs",
        (
            ([(CatBoostModel, cb_test_params)] if CB_AVAILABLE else [])
            + ([(XGBModel, xgb_test_params)] if XGB_AVAILABLE else [])
        )
    )

Then you have to slightly adapt the code below to use the new parameters and setup the model dynamically.

[dennisbader]

the same here as above, let's cover both models in the same test via parametrization

[dennisbader]

Looks great now, thanks a lot for the updates @ozink-u8 🚀

I pushed some minor changes, now everything is good to go 💯