provide example code for suggestions in misc/warnings

[petrelharp]

Summary:

Some of the suggestions in the (very nice!!) page provided to explain convergence warnings](https://mc-stan.org/misc/warnings.html) are hard to figure out how to actually implement. It would help greatly to have some example code (and maybe additional methods in rstan?). There's a bunch of great work underlying these convergence diagnostics, and a bit more documentation would help make them much more accessible.

Description:

The linked page, https://mc-stan.org/misc/warnings.html , is the one that shows up in the warning message itself, so probably quite a few people visit it. Here are the suggestions that are hard to figure out:

BFMI low:
Look at the pairs plot to see which primitive parameters are correlated with the energy__ margin.

As in #719, one can't put "energy__" directly into the pars argument to pairs; one puts it into condition, but that's not so obvious (and, hard to see).

R-hat:
Look at Bulk- and Tail-ESS for further information.

There are ess_bulk( ) and ess_tail( ) functions, but they don't take stanfit objects; it seems like what one wants to do is, say

apply(extract(fit, permuted=FALSE), 3, ess_bulk)

... ?

Look at the rank plot to see how the chains differ from each other.
Look at the local and quantile efficiency plots.

It looks like that's what's happening in this bayesplot issue; so maybe a link to the bayesplot diagnostics vignette is in order?

There's a few more suggestions to look at these things.

I'm posting this here because the URL is in the output of rstan::sampling( ), but maybe it should be elsewhere, since really it's part of mc-stan.org.

Thanks for a great tool!!!