add tests of frz posterior with continue cascade

[ntemiq]

This PR includes a minimal example of using freeze_type posterior in a workflow with continue_cascade - this mimics the process used by our cluster, breaking up fits into smaller groups and calling continue cascade

It appears that calling a fit_database below the root node in continue cascade is leading to drift in the mulcov priors that are frozen with freeze_type==posterior.

[bradbell]

I moved this to a test; see
https://github.com/bradbell/at_cascade/blob/master/test/csv/freeze_posterior.py
Note that I commented out the documentation by changing {xrst_begin -> {xrst_comment.

You can see the corresponding bug fix in
5222d68

If you want to make this an example, merge the current master into this pull request, move the test to the example directory, and test the corresponding documentation using
https://github.com/bradbell/at_cascade/blob/master/bin/run_xrst.sh

You should also run check_all.sh when you think you have a working version.