Add target components logic to pipelines

[alex-hse-repository]

🚀 Feature Request

Add target components logic to pipelines to the base classes of our pipelines

Proposal

1. In all the classes from folders ensembles, pipeline:

• In methods _forecast, _predict add flag

return_components: bool

• In methods forecast, predict add flag

return_components: bool = False

• Add NotImplemented error for the case return_components=True

2. In BasePipeline pass return_components flag in forecast, predict into _forecast, _predict
3. In ModelPipelinePredictMixin pass return_components flag in _predict into the model's predict method
4. In Pipeline pass return_components flag in the necessary places inside the forecast and _forecast

Test cases

Run this test cases for dummy implementation of NonPredictionIntervalContextIgnorantAbstractModel, NonPredictionIntervalContextRequiredAbstractModel, PredictionIntervalContextIgnorantAbstractModel, PredictionIntervalContextRequiredAbstractModel. There should be different implementation for forecast and predict (i.e. return different constants).

1. For Pipeline check that when return_components=True dataset contain the correct target components for methods forecast, predict for dummy implementations

2. For all the other pipelines check that they throw Notimplemented on return_components=True

Additional context

No response