Migration fixes

[richafrank]

A few unrelated improvements

[ssanderson]

Is this just broken in the current incarnation? I thought we had tests for this case...

[adzuci]

I see tests for renaming directories, but I couldn't find tests on how checkpoints interact with things:
https://github.com/quantopian/pgcontents/blob/master/pgcontents/tests/test_pgmanager.py#L124

I'm not quite sure what the test logic would be either & tried to find what actually calls rename_checkpoint and couldn't figure out where it would be:
https://github.com/jupyter/notebook/search?utf8=%E2%9C%93&q=rename_checkpoint

EDIT: Nevermind, I just fail at searching, rename_all_checkpoints calls it and is called by rename: https://github.com/jupyter/notebook/blob/a6c5be5d60f122400b65d9c8d6488e78f25b9c2f/notebook/services/contents/manager.py#L231

[ssanderson]

@adzuci it gets called here: https://github.com/jupyter/notebook/blob/a6c5be5d60f122400b65d9c8d6488e78f25b9c2f/notebook/services/contents/manager.py#L234. That should be exercised by https://github.com/ipython/ipython/blob/3.x/IPython/html/services/contents/tests/test_contents_api.py#L511, which is executed in our test suite for sure.

[richafrank]

rename_checkpoint is only called by rename_all_checkpoints in the base class, but PostgresCheckpoints overrides rename_all_checkpoints to rename all the checkpoints in a single SQL statement, instead of renaming them one by one. I'm going to replace this fix with raise NotImplementedError() to show that it's not currently expected that something would call it.