Useful plots for MRP

@jgabry and I were talking that it might be handy to have extra functionality in bayesplot to do MRP style plots.

Currently you can use BayesPlot with posterior_predict objects provided the observed and newdata are of the same size. This is useful to be able to compare observed versus predicted (when using posterior_predict) to do model diagnostics.

For MRP it would be useful to increase this functionality where newdata of a different size to the fitted data, and also to include functionality to add a weighted estimate (if weights are available), a raw sample estimate and the full posterior predictive estimate.

I know @bbbales2 has some great plots for MRP, so he might be able to weigh in. I'll follow up with an example to show what I envision it would look like. @mitzimorris and @andrewgelman might be interested too.

Start by reviewing bayesplot's existing posterior_predict plotting functionality and the MRP plotting examples mentioned in the issue. The work would be complete when plots support newdata of a different size from the fitted data and can show weighted, raw-sample, and full posterior-predictive estimates.