Evaluate() to return aggregated metric evaluation across unique_ids.

[tackes]

Description

It would be a nice feature to be able to calculate metrics for global models, instead of only by unique_id.
IE if we have an ML models (say Catboost and LGB) with multiple forecasts for each unique_id, I would like to know the MAE for Catboost vs LGB across all unique_ids.

Use case

No response