fix(metrics): name the wrapper mock correctly and document NaN r2 with one valid pair (SIENTIAPDE-2215)

model_manager/utils/repository/data_manager_repository.py

@@ -358,7 +358,7 @@ class DataManagerRepository(SientiaMonitoring):
         behavior, assuming that predictions (y_pred/y_train_pred) are already on the
         correct scale for metric calculation (any scaling is handled inside the
         model wrapper). Pairs with a NaN target or a NaN prediction are left out of
-        mse/mae/r2.
+        mse/mae/r2. With a single valid pair left, r2_val is NaN (R² needs 2 samples).
 
         Args:
             tmr: Training result containing: