Add validation in training

[jakub-piw]

Validation set needs to be created in the begining of training from original set For greedy approach:

• Performance on validation set will be decisive in the selection of model in iteration.
• Fitting is stopped when metric on validation set reaches defined level. For sequential approach:
• Fitting is stopped when metric on validation set reaches defined level or thera are no models left.