Add 2d density plots as alternative to scatterplot

Allow using a density plot instead of a ``discrete_scatter`` inside of ``plot`` (in ``plot_classification_continuous``)

We might even use seaborn:
https://seaborn.pydata.org/tutorial/distributions.html

Or maybe we have to use kdes ourselves. There's also bubble sets:
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=5290706

And several cool ideas here that might be worth exploring:
https://www.ncbi.nlm.nih.gov/pubmed/30136987

I like the voronoi with the bar charts in it for example.

Start at plot_classification_continuous and trace how plot currently uses discrete_scatter. Review the linked seaborn distributions tutorial and the other cited visualization ideas before narrowing the design. Done means plot supports a defined 2D density alternative to the scatterplot path.

Written by the indexing model from the issue text.