feat: Compare CrossValidationReports

[auguste-probabl]

This PR makes it possible to pass a list of CrossValidationReports to the ComparisonReport (which currently only accepts EstimatorReports).

For now, plots like Comparison.metrics.roc_curve are not supported for ComparisonReports holding CrossValidationReports.

---

Example:

>>> from sklearn.datasets import make_classification
>>> from sklearn.linear_model import LogisticRegression
>>> from skore import ComparisonReport, CrossValidationReport
>>> X, y = make_classification(random_state=42)
>>> estimator_1 = LogisticRegression()
>>> estimator_2 = LogisticRegression(C=2)  # Different regularization
>>> report_1 = CrossValidationReport(estimator_1, X, y)
>>> report_2 = CrossValidationReport(estimator_2, X, y)
>>> report = ComparisonReport({"model1": report_1, "model2": report_2})
>>> report.metrics.report_metrics(aggregate=None)
                                                    Value
Metric       Label / Average Estimator Split
Precision    0               model1    Split #0  1.000000
                                       Split #1  1.000000
                                       Split #2  1.000000
                                       Split #3  1.000000
                                       Split #4  0.909091
...                                                   ...
Predict time                 model2    Split #0  0.000185
                                       Split #1  0.000177
                                       Split #2  0.000187
                                       Split #3  0.000177
                                       Split #4  0.000188

[80 rows x 1 columns]
>>> report.metrics.report_metrics()
                                            mean       std
Metric       Label / Average Estimator
Precision    0               model1     0.981818  0.040656
                             model2     1.000000  0.000000
             1               model1     0.981818  0.040656
                             model2     0.981818  0.040656
Recall       0               model1     0.980000  0.044721
                             model2     0.980000  0.044721
             1               model1     0.980000  0.044721
                             model2     1.000000  0.000000
ROC AUC                      model1     0.996000  0.008944
                             model2     0.996000  0.008944
Brier score                  model1     0.022532  0.013130
                             model2     0.018731  0.014566
Fit time                     model1     0.001959  0.000559
                             model2     0.002603  0.000185
Predict time                 model1     0.000198  0.000014
                             model2     0.000183  0.000005

---

Closes #1414

Todo:

• Coverage
• mypy
• Docs probably
• Fix bug with progress bars not getting cleaned up
• Format tables properly according to feat(ComparisonReport): Be able to compare several CrossValidationReport #1414
• Make plots work deferred
  • Add NotImplementedError to plots
• Test what happens when aggregate is passed to ComparisonReport[EstimatorReport]
• Test what happens when X, y is passed to ComparisonReport[CrossValidationReport]
• Add doctests to ComparisonReport showcasing ComparisonReport[CrossValidationReport]
• Add timings
• Check if some estimators are named the same, if so make them different (do it when initializing report.report_names_)

[github-actions]

Documentation preview @ 19b6926

[auguste-probabl]

Discussed IRL: The code common between ComparisonReport and CrossValidationComparisonReport should be factored in a BaseComparisonReport.
It makes more sense for us to have just ComparisonReport, that can take either EstimatorReports or CVReports.

[auguste-probabl]

This new constraint results from #1536; indeed, if the same report is passed twice, when we reset the progress bar at the end of the first CVReport computation (see below), when the second CVReport computation starts, the progress object is set to None, whereas it should be set to the ComparisonReport's progress object.

skore/skore/src/skore/utils/_progress_bar.py

Lines 85 to 86 in 836f171

	self_obj._parent_progress = None
	self_obj._progress_info = None

[auguste-probabl]

Note that right now this constraint is only really useful when comparing CVReports, not when comparing EstimatorReports. I didn't put it in the if-block because I can imagine that one day EstimatorReports might have more progress bars of their own.

[auguste-probabl]

This is necessary because of the new constraint that all reports must be distinct

[glemaitre]

Let's merge now. @auguste-probabl could you open an issue to track the improvement/refactoring for the progress bar to not forget.