Idea: Outlier handling on X scale

[mayer79]

Could we introduce an option for plot_marginal() and maybe for some other plotting functions to winsorize outliers in numerical x before calculating bins?

There are a couple of options, and all have drawbacks:

1. Winsorize values outside [q1 - alpha * IQR, q3 + alpha * IQR], where qi is the i-th quartile: Expensive to calculate on large data, but one could subsample on 10k rows.
2. Use mean(x) +- alpha std(x). Drawback: mean and std are not robust.
3. Set manual limits. Drawback: User has to work.

Values outside the limits would fall into the most extreme bins for calculations.