CCrossValidationSubProblem in public interfaces

[AlekseyLebedev]

Hello.

I want to ask you to put CCrossValidationSubProblem into public interfaces (from https://github.com/neoml-lib/neoml/blob/master/NeoML/src/TraditionalML/CrossValidationSubProblem.h).
It would be useful for custom cross-validation algorithms implimentations.

For nowdays, I have copy-paste of that code in another repo.

[AlekseyLebedev]

Another solution of my case is to add those subproblems to

// Cross-validation result
struct NEOML_API CCrossValidationResult {
	CPtr<const IProblem> Problem; // the data on which cross-validation was performed
	CObjectArray<IModel> Models; // the models trained
	CArray<double> Success; // the classification quality for each of the models on the corresponding training set
	CArray<CClassificationResult> Results; // the vector classification results
	CArray<int> ModelIndex; // the index of the model that classified the given vector
};

neoml/NeoML/include/NeoML/TraditionalML/CrossValidation.h

Line 24 in bb8d684

// Cross-validation result

My scenario is to find optimal tree number in gradient boosting by crossvaludaion.
To do that, I use ClassifyEx method of ClassifyEx of IGradientBoostModel
When I use CCrossValidation I got all models and al their classification results. However, it is not ClassifyEx results of IGradientBoostModel. Moreover, subproblems used to train those models are unaccesable.

If I have test subproblems in CCrossValidationResult I could solve my case. I suppose, test and train subproblems in result of cross-validation can be useful for others.