Add VBA functionality

[sjsrey]

Following the suggestions raised in #36 by @slumnitz.

[darribas]

How would this work exactly? Just opening the discussion.

My main concern is in making intuitive the classification of the vba variable given the classification of the main variable is obvious since this will be a method of a given classifier. For example given a continuous variable y:

classi = Quantiles (y)
classi.plot(geoms)

This would be a standard choropleth. Now, if I have another variable (e.g. confidence) that I want to use to modulate alpha values. How should this be passed in? More importantly, how should we allow the user to determine the classiffication scheme for that? A couple of ideas (totally brainstorming so might not make sense):

1. We use the style of choropleth_vba and allow the user to pass any classifier. This is consistent but a bit awkward in a way since they would pick the classifier for y in one way and for confidence in another)
2. We don't allow the user to pass classifiers for vba and use the same classifier and same parameters for confidence as used for y. This is consistent within the method and simpler to set up but arguably a bit too restrictive.
3. Something in the middle?

[sjsrey]

Another possibility:

class_a = Quantiles(a)
class_b = Quantiles(b)
vba.plot(class_a, class_b, geoms)

where vba is an instance of a new class.

I think loading up the plot method in the univariate classifiers is going to complicated things.

[darribas]

I really like that approach. I'd maybe call it choropleth instead of vba?

[slumnitz]

splot.vba_chorpleth actually does exactly this under the hood. For maintaining purposes we might want to consider to leverage functionality for both and just store it at one place.

[martinfleis]

Can this be closed since the plotting is in splot?

[jGaboardi]

I think so. We can reopen if needed.