Docs add example of Resemblance model trained on X and the target y

[Matgrb]

Typically when you train the Resemblance model, you provide two data samples X1 and X2, without the labels y1 and y2. This allows you to analyse how the relations between features differ between the two samples.

However, this disregards the relations between the features and the target as well as the shift of the target. One way to tackle that, is including y into one of the columns of X.

It would be helpful to add to the resemblance model docs a section presenting this use case and how to analyse the results and compare them with resemblance model trained on X1 and X2 only.