[DEE-456] nlp dummy model refactoring

[yromanyshyn]

@Nadav-Barak

Currently we have a dummy model for tabular and another one for nlp but they basically do the same stuff.
We would like them to have common abstract containing most of the logic with child class implement datadype specific functionalities.

they do not have much in common, in general, _predict and _predict_proba methods could be separated into some kind of base class, but in this case, we would need to unify nlp.TextData and tabular.Dataset APIs, everything else is just a validation logic that is pretty specific IMO. So, I do not see a need in creating some kind of ambiguous and not clear abstractions just to reduce a few lines of code

In general, I just refactored the nlp dummy model and moved the validation logic into the input_validation.py module

[Nadav-Barak]

To verify, either way predictions and probabilities return as np array?

[yromanyshyn]

@Nadav-Barak

To verify, either way predictions and probabilities return as np array?

yeap