Fix sklearn tree importances retrieval bug (#377)

When obtaining the tree feature kpis we access the JSON report path
`treePreparationReport.variablesStatistics`. If there are informative
variables but no informative tree `variablesStatistics` is not defined.
We didn't check that and it lead to a bug in this rare case.

Author: folmos-at-orange

CHANGELOG.md

@@ -11,6 +11,9 @@
 ### Added
 - (`sklearn`) Support for boolean and float targets in `KhiopsClassifier`.
 
+### Fixed
+- (`sklearn`) Crash when there were no informative trees in predictors.
+
 ## 10.3.0.0 - 2025-02-10
 
 ### Fixed

khiops/sklearn/estimators.py

@@ -1674,10 +1674,14 @@ def _fit_training_post_process(self, ds):
         else:
             pair_feature_evaluated_names_ = []
             pair_feature_evaluated_levels_ = []
-        if "treePreparationReport" in self.model_report_raw_:
-            tree_preparation_report = self.model_report_raw_["treePreparationReport"][
-                "variablesStatistics"
-            ]
+        if (
+            "treePreparationReport" in self.model_report_.json_data
+            and "variablesStatistics"
+            in self.model_report_.json_data["treePreparationReport"]
+        ):
+            tree_preparation_report = self.model_report_.json_data[
+                "treePreparationReport"
+            ]["variablesStatistics"]
             tree_feature_evaluated_names_ = [
                 tree_preparation_report[i]["name"]
                 for i in range(0, len(tree_preparation_report))