Clean up brmsfit objects

[paul-buerkner]

Currently, brmsfit objects store some information multiple times at different places to retain backwards compatibility with older versions of brms. In version 3.0 (still a long way to go), we should take the opportunity to clean that up. This will include the following changes:

• Update formula following the changes in Restructure brmsformula objects #476.
• remove elements family and autocor as they are natively stored inside the formula element.
• remove element ranef as it can be easily computed from formula and data without much computational overhead.
• make data.name an attribute of data.
• remove deprecated element stan_funs.

[strengejacke]

Just a quick thought: Do you really want to remove ranef()? I think this is a convenient shortcut, and it helps other people writing generic functions that work for brms, lme4 or glmmTMB.

[paul-buerkner]

You may be right. I don't like updating it for old objects whenever I add a new multilevel feature, but I think this is acceptable if its helpful for users or developers.

[paul-buerkner]

@strengejacke Did you understood my OP as if I wanted to remove the ranef() method? That's not what I ment. Instead I talked about the $ranef element of brmsfit objects which contains information about the "random" effects structure not posterior samples for the corresponding parameters.

[strengejacke]

Ok, yes, than I misunderstood you in first place.