Create hyperparameters configs for pipelines and adapt pipelines to use configs

[ktagowski]

We should provide more clearer way how to pass hyperparameters for pipeline (currently we use multiple kwargs).

Something like HPS ConfigSpace

class ClassificationConfig:
    max_epochs: int = 30
    learning_rate: float = 0.001
    ...

    def parse_params():
        pass

And then we should adapt it to our pipelines