Describe how `brms` implements models for censored data

[wzbillings]

By extracting the Stan code from fitted brms models, we should be able to determine how brms translates its formula language into likelihood models for censored outcomes (since the documentation never says how it works, we have to reverse engineer it or ask Burker/the stan forum for details, I guess). Hopefully the answer is just by updating the likelihood in a way that we can understand.

Once we can figure out the extract Stan code, we should do a writeup explaining what a brms formula for censored data actually means in likelihood model terms.