Stacked-area plot for ensemble model contributions

[egordm]

As a forecaster working with ensembles
I want to see how much each base model contributes to the forecast at each timestep
So that I can explain and debug ensemble behaviour

🌍 Background

EnsembleForecastingModel.predict_contributions() returns each base model's contribution per
quantile per timestep as a TimeSeriesDataset (columns {model}__{quantile}, e.g. gblinear__P50).
There is no model-level view for it yet. ContributionsPlotter (openstef_models.explainability.plotters)
already does the per-feature version; this is its model-level sibling.

Important

Changes should be done by forking main branch (main) and then make a PR back onto main branch.
Please read the contributing guide before you start.

❗Priority

Low. Explainability convenience, no deadline.

Definition of Done:

✅ Acceptance criteria

• A new view takes the predict_contributions() TimeSeriesDataset and returns a Plotly go.Figure, following ContributionsPlotter conventions.
• Two-panel, shared x-axis like ContributionsPlotter.plot_heatmap: top is the forecast over time, bottom is a stacked area of each base model's contribution.
• The view takes a quantile argument (default P50) selecting which quantile to plot.
• Can be saved to a standalone HTML file.
• Demonstrated in an example (e.g. examples/tutorials/model_explainability.py).

📄 Documentation criteria:
Add it to the model explainability tutorial.

🧪 Test criteria:
Smoke test that the plot builder returns a go.Figure from a small synthetic contributions dataset.

⌛ Dependencies:
N/A

🚀 Releasing:
N/A

Other information:
Prior work: #789 (closed) implemented this against the outdated release/v4.0.0 branch. Useful
reference for the plotting approach (per-model series, quantile buttons, HTML export).