feat(skore/EstimatorReport)!: Move arguments from .summarize() to .frame() (probabl-ai#2536)

<!--
🙌 Thanks for contributing a pull request!

If this is your first contribution, take a look at our contribution
guidelines:
https://docs.skore.probabl.ai/dev/contributing.html
-->
#### Change description
<!--
Please describe what your contribution changes for skore.

Here is some inspiration for what to write here:
- Are you adding a new feature? Fixing a bug?
- Can you give an example of your change in action, e.g. a snippet of
code or a plot?
- Is your change likely to break users' code?
- Are there any other details the reviewer should be aware of, such as
API design choices, performance characteristics or edge cases?

Please reference issues/PRs when possible, e.g. "Fixes probabl-ai#1234", "Closes
#3456", "See also #7890".
More information
[here](https://github.com/blog/1506-closing-issues-via-pull-requests).
-->

`EstimatorReport.metrics.summarize()` no longer accepts arguments
`flat_index` or `favorability`. These have been moved to
`MetricsSummaryDisplay.frame()`.
This moves the responsibility of displaying things from
`EstimatorReport._MetricsAccessor` to `MetricsSummaryDisplay`.

```python
# Before
report.metrics.summarize(flat_index=True, favorability=True).frame()

# After
report.metrics.summarize().frame(flat_index=True, favorability=True)
```

This is a breaking change.

The rest of the PR consists in various refactorings, in particular the
tests have been updated to reflect the change in responsibility:
- Tests of `summarize()` were extracted from
`tests/unit/reports/estimator/metrics/test_numeric.py` to a new file,
`tests/unit/reports/estimator/metrics/test_summarize.py`.
- `test_summarize.py` now tests that `summarize()` behaves well, and
that the output of `summarize()` has a well-formed DataFrame. `.frame()`
is never called.
- `displays/metrics_summary/test_estimator.py` has been rewritten to
specifically test `.frame()` arguments.

A number of commits are included that could be pulled out into a
separate PR if needed.

Closes probabl-ai#2533
Supersedes probabl-ai#1839 in part

#### Contribution checklist
<!--
Below are some of the criteria that the review will include.
Feel free to use it as a checklist to ensure that your contribution is
high-quality.
-->

- [x] Unit tests were added or updated (if necessary)
- [x] Documentation was added or updated (if necessary)

#### AI usage disclosure
<!--
If AI tools were involved in creating this PR, please check all boxes
that apply
below and make sure you understand our [automated contributions
policy](https://docs.skore.probabl.ai/dev/contributing.html#automated-contributions-policy)
-->
AI tools were involved for:
- [ ] Code generation (e.g., when writing an implementation or fixing a
bug)
- [x] Test/benchmark generation
- [ ] Documentation (including examples)
- [ ] Research and understanding

In particular I used Claude to increase coverage.
<!-- Any other comments can go here. Thanks again for contributing! -->

Author: auguste-probabl

examples/use_cases/plot_employee_salaries.py

@@ -156,7 +156,7 @@
 
 # %%
 hgbt_split_1 = hgbt_model_report.estimator_reports_[0]
-hgbt_split_1.metrics.summarize(favorability=True).frame()
+hgbt_split_1.metrics.summarize().frame(favorability=True)
 
 # %%
 # The favorability of each metric indicates whether the metric is better

skore/src/skore/_sklearn/_comparison/metrics_accessor.py

@@ -169,7 +169,9 @@ class is set to the one provided when creating the report. If `None`,
                 results.index = results.index.str.replace(
                     r"\((.*)\)$", r"\1", regex=True
                 )
-        return MetricsSummaryDisplay(results)
+        return MetricsSummaryDisplay(
+            data=results, report_type=self._parent._report_type
+        )
 
     def _compute_metric_scores(
         self,
@@ -199,15 +201,32 @@ def _compute_metric_scores(
                 )
             )
 
-            kwargs = dict(
-                data_source=data_source,
-                **metric_kwargs,
-            )
+            kwargs = dict(data_source=data_source, **metric_kwargs)
             if is_cv_report:
                 kwargs["aggregate"] = None
 
+            # FIXME(#1837)
+            # "favorability" and "flat_index" are passed to `.frame()` for
+            # EstimatorReports while they are passed to `.summarize()` for
+            # CrossValidationReports
+            frame_kwargs = {}
+            if not is_cv_report:
+                for key in ("favorability", "flat_index"):
+                    if key in kwargs:
+                        frame_kwargs[key] = kwargs.pop(key)
+
+            if "favorability" in kwargs:
+                favorability = kwargs["favorability"]
+            elif "favorability" in frame_kwargs:
+                favorability = frame_kwargs["favorability"]
+            else:
+                favorability = False
+            favorability = cast(bool, favorability)
+
             individual_results = [
-                result.frame() if report_metric_name == "summarize" else result
+                result.frame(**frame_kwargs)
+                if report_metric_name == "summarize"
+                else result
                 for result in track(
                     parallel(
                         joblib.delayed(getattr(report.metrics, report_metric_name))(
@@ -224,14 +243,14 @@ def _compute_metric_scores(
                 results = _combine_estimator_results(
                     individual_results,
                     estimator_names=self._parent.reports_.keys(),
-                    favorability=metric_kwargs.get("favorability", False),
+                    favorability=favorability,
                     data_source=data_source,
                 )
             else:  # "CrossValidationReport"
                 results = _combine_cross_validation_results(
                     individual_results,
                     estimator_names=self._parent.reports_.keys(),
-                    favorability=metric_kwargs.get("favorability", False),
+                    favorability=favorability,
                     aggregate=aggregate,
                 )
 

skore/src/skore/_sklearn/_cross_validation/metrics_accessor.py

@@ -168,7 +168,7 @@ class is set to the one provided when creating the report. If `None`,
                 results.index = results.index.str.replace(
                     r"\((.*)\)$", r"\1", regex=True
                 )
-        return MetricsSummaryDisplay(summarize_data=results)
+        return MetricsSummaryDisplay(data=results, report_type="cross-validation")
 
     def _compute_metric_scores(
         self,
@@ -196,8 +196,15 @@ def _compute_metric_scores(
                 )
             )
 
+            frame_kwargs = {}
+            for key in ("favorability", "flat_index"):
+                if key in metric_kwargs:
+                    frame_kwargs[key] = metric_kwargs.pop(key)
+
             results = [
-                result.frame() if report_metric_name == "summarize" else result
+                result.frame(**frame_kwargs)
+                if report_metric_name == "summarize"
+                else result
                 for result in track(
                     parallel(
                         delayed(getattr(report.metrics, report_metric_name))(
@@ -220,7 +227,7 @@ def _compute_metric_scores(
             # Pop the favorability column if it exists, to:
             # - not use it in the aggregate operation
             # - later to only report a single column and not by split columns
-            if metric_kwargs.get("favorability", False):
+            if frame_kwargs.get("favorability", False):
                 favorability = results.pop("Favorability").iloc[:, 0]
             else:
                 favorability = None