Find a way to merge some of C-TMLE and CV-TMLE internals

[olivierlabayle]

This is in order to simplify the development of new features that require to be propagated throughout all the codebase (e.g. prevalence weighting).

For instance:

• Both require cross-validation of ML algorithms but the way they use them is slightly different, for now we have two disjoint implementations which is not ideal.
• ...