Improve regressors handling in TSDataset

[alex-hse-repository]

🚀 Feature Request

Make make_future and train_test_split pass the list of regressors to the generated datasets.

Motivation

In #327 we added parameter known_future in the constructor of TSDataset. However methods make_future and train_test_split don't use this parameter constructing the output.

Proposal

1. Pass self.regressors to the constructor in make_future here
2. Pass self.regressors to the constructor in train_test_split here and here
3. Create test for this cases

Test cases

1. Check that future part inherit the list of regressor from dataset after applying make_future
2. Check that train and test parts inherit the list of regressor from dataset after applying train_test_split

Alternatives

No response

Additional context

No response

Checklist

• I discussed this issue with ETNA Team