add stratification post

[giuseppec]

This post covers three ways to achieve reliable resampling:

1. Stratified resampling for classification problems where each train/test split maintains the target class distribution of the original data set.
2. Blocked resampling where a grouping factor determines which observations should be together in train/test splits.
3. Using predefined and manually created folds for the train/test splits.

[missuse]

Ready from your side @giuseppec @missuse?

I don't see any problems. Although, I must admit, I usually see many problems only after I click submit/merge/publish