Dataset `stats` function should include descriptive statistics about the data and targets.

[HesitantlyHuman]

Currently the stats function is quite lackluster, ideally, for things like strings we should see stats about length, for things like class labels proportions and for regression targets, stuff like averages and distributions.