feature: Added support for evaluation relative to a single date. Useful for production usecases.

[egordm]

This pull request introduces support for reference-date-based windowed evaluation in the OpenSTEF Beam evaluation pipeline. The main goal is to allow users to compute evaluation metrics focused on a specific time window around a given reference date, rather than always evaluating over all possible windows. This is achieved by propagating a new reference_date parameter through the pipeline and updating window iteration logic accordingly. Unit tests are also added to ensure correct behavior.

Enhancements to windowed evaluation:

• Added a reference_date parameter to the evaluation pipeline (evaluation_pipeline.py), enabling evaluation focused on a specific time period around the provided date. This parameter is propagated through run, run_for_subset, and _evaluate_subset methods, and is documented in their docstrings. [1] [2] [3] [4] [5] [6] [7] [8]
• Updated the window iteration utilities (window_iterators.py) to accept and handle the reference_date parameter. When provided, only a single window corresponding to the reference date is yielded, rather than all possible windows. [1] [2] [3] [4] [5]

Window model improvements:

• Added get_timerange and is_future methods to the Window class, which calculate the time range for a window based on a reference date and check if a window ends after the reference date, respectively.

Testing:

• Introduced new unit tests in test_window_iterators.py to verify that window iteration returns either all windows or a single window depending on whether reference_date is set.

Minor updates:

• Updated imports to include datetime where needed for the new functionality. [1] [2]

[MvLieshout]

LGTM!