fix: Ensemble Forecasting Model postprocessing fit and transform

[MvLieshout]

This pull request enhances the flexibility and clarity of the ensemble forecasting workflow by introducing more granular postprocessing steps for both individual forecasters and the combiner. It also improves code organization and type safety in the workflow configuration. The most important changes are:

Model Architecture and Processing Pipeline:

• Added model_specific_postprocessing and combiner_postprocessing fields to the EnsembleForecastingModel to allow distinct postprocessing steps for each forecaster and for the combiner, in addition to the existing shared postprocessing pipeline.
• Updated the _predict_forecaster and _predict_combiner methods to apply the new postprocessing pipelines at the appropriate stages, ensuring that predictions are transformed as intended before being passed to the combiner and before final output. [1] [2]
• Modified the _fit_combiner method to fit the combiner postprocessing pipeline on training predictions, ensuring proper transformation during training.

Workflow Configuration Improvements:

• Refactored the create_ensemble_forecasting_workflow function to construct and inject the new postprocessing pipelines (common_postprocessing, model_specific_postprocessing, and combiner_postprocessing) into the model, replacing the previous monolithic postprocessing approach. [1] [2]
• Improved type safety and code clarity in the workflow configuration by using TYPE_CHECKING for type hints and reorganizing imports. [1] [2]

These changes provide greater modularity and control over the preprocessing and postprocessing steps in ensemble forecasting workflows, making it easier to customize and extend the pipeline for different use cases.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[egordm]

LGTM!