Allow analysis.R to instantiate a full brmsfit object

[WardBrian]

With #323 we can generate stan code/data from brms, but we do this with empty=TRUE.

In analysis.R, it would be nice if we could actually insert the draws from SP back in. This requires creating a stanfit object, which can be created from the csv files output by cmdstan (using the brms::read_csv_as_stanfit function), but may be harder to build ourselves.

We'd also want to check that the brm call in analysis.R is for the same model as the one in data.R was if we supported this

Originally posted by @wds15 in #202

[WardBrian]

Serendipitously, someone on the Stan forums has written another stanfit-from-nothing function:
https://discourse.mc-stan.org/t/bayesian-bootstrap-implementation-for-brms/40522?u=wardbrian

Between this and brms::read_csv_as_stanfit I suspect we could cobble something together.