ImbalanceComparator should also check imbalance compared to entire dataset

[kwinkunks]

It would be interesting to compare the imbalance in train and in test to the imbalance in the combined train+test dataset, to make sure they are more or less the same.

E.g. imagine both train and test were imbalanced, but in a different way (a) to each other and (b) to the combined dataset.

Storing the histogram should enable this, if we don't already do it.