Make mcmc_pairs() return a ggplot object instead of a bayesplot_group object (Fix #268)

[billdenney]

This PR makes mcmc_pairs() return a ggplot object instead of a bayesplot_group object (Fix #268).

This is a DRAFT PR for review and comment!!! The reason that it is a draft is that it emits all sorts of warnings that would need to be fixed before merging, but I know that there are strong preferences about the graphical organization of the output of mcmc_pairs(), and I wanted to provide this draft to allow for testing and review. For a final PR, I would ensure that warnings were not present (or that they were minimized as much as possible).

I believe that I captured all the major functionality from the existing mcmc_pairs() output, but there were some interactions around how divergent transitions, max tree depth hits, and hex plots that may be off.

[billdenney]

Some differences that I think are notable are:

1. x-axes and y-axes are aligned now (because they are facets)
2. row and column titles are now in strip labels instead of subtitles of the diagonal elements (again because they are facets)
3. the vertical scale for the density or histogram plots along the diagonal are the full height of the facet rather than sharing the height with the subtitle

[jgabry]

Hey @billdenney thank you for this and sorry for the slow reply! I'm caught up in a bunch of other stuff at the moment but hopefully I'll have time to give you some comments soon.