Created by: Dreamlone
Within this PR, the functionality of modeling different targets in one chain is prepared. Decomposition operations are also implemented for classification and regression tasks.
For example, chain for modeling error for the multiclassification task can be looking like this: