Increase support for circular and discrete variables in ArviZ

This Google Summer of Code 2020 project aims to increase support for circular and discrete variables in the ArviZ library. ArviZ currently provides tools and diagnosis for Bayesian data analysis and it is very well supported for continuous variables. However, there is plenty of space for new developments when it comes to discrete and circular variables. As a result of this project, several new plots will be incorporated into ArviZ, for example: mosaic plot, agreement plot, circular KDE, circular histogram. On the other hand, ArviZ current implementations will be revised and extended for the circular and discrete variable setting.