posteriordb fork overhaul

[fonnesbeck]

This is more a starting point for discussion than anything. Here, is what I've done:

• stripped out all of the R-specific code
• integrated the API tools from posteriordb-python.
• added benchmarking via asv
• ported the prophet model and the non-centered eight schools model as test cases

Whether any of this is any good depends on whether we agree on what our posteriordb fork is going to be used for. I'm assuming we don't want to maintain R code, we want to carry the Stan models, and we want to use it for PyMC/PyTensor/nutpie benchmarking.

[twiecki]

Should we just merge and then continue building on this?