Add new vi plots

[aloctavodia]

Computation of variable importance is splitted into one function that computes the importance and two that plots the results.

For example.

vi_results = pmb.compute_variable_importance(idata_bikes, μ_, X);

vi_results is a dictionary, we can do

pmb.plot_variable_importance(vi_results, X);

and/or

pmb.plot_scatter_submodels(vi_results, grid=(2,2))

This last function needs more love, for instance, we could have error bars per observation, as we did for Friedman's example, and we need to improve labels, aesthetics, etc.

There is also a function to plot the variable inclusion

pmb.plot_variable_inclusion(idata_bikes, X)

This could be used as a guide for setting the parameters for the backward_VI method