This line has seen some action before in I1867f633, and before that in
I265ef862. Rereading those commit messages, I see no reason why we
should be opening the repo and possibly triggering auto-rebuilding when
NoteDb writes and reads are both disabled.
This was causing NoSuchChangeExceptions in the case where the Change
contained a noteDbState field but the ref was absent. That logic is also
questionable, for reasons now mentioned in a TODO. But really we should
never be reaching the openHandle method when NoteDb is completely off.
Add a regression test that would have caught this.
Change-Id: If0970e1cf61e1d98ccbb3ce27549186f5771466a
c893c02b32