Empty node titles in content overview

Patch followed :-) Hopefully there are enough comments in this to make it clear what is going on! From the Doxygen:

Language hierarchy enforces that languages only fallback to their parent
* languages, whereas Drupal core would allow all languages to be fallbacks.
* Title replacement fields are a special case, since we may want to show a
* title (e.g. to admins) even if there are no translations in a parent language
* so all languages are allowed as fallbacks for those. We give primacy to the
* entity's own language first, then parent languages ahead of cousin languages.
* Since this may mean content is exposed that would not normally be accessible
* to users, a permission is used to toggle this functionality.

Two points of note, from a real-world discussion:

* entity_language() is not used for the final fallback case, which means different behaviour when there is a form language (with entity translation). Is that correct?
* It's actually possible for that final fallback, entity_language(), to return a language that there are not values for, since that might just return the site default language. So we should probably do a final fallback that loops through any remaining languages that haven't been checked, in weight order, until a value is actually found.
* It might even be worth doing an early quick edge case test, if there are no values in any languages, or just one, before the main bulk of the code tries to proceed.

* entity_language() is not used for the final fallback case, which means different behaviour when there is a form language (with entity translation). Is that correct?

This won't matter either way, as form language won't interfere at the point when it gets used. I could believe there is some obscure case when it does matter, but I think let's go with the simpler

* It's actually possible for that final fallback, entity_language(), to return a language that there are not values for, since that might just return the site default language. So we should probably do a final fallback that loops through any remaining languages that haven't been checked, in weight order, until a value is actually found.

My mistake, the patch does do this already.

* It might even be worth doing an early quick edge case test, if there are no values in any languages, or just one, before the main bulk of the code tries to proceed.

This point would still be valid.

New test dependencies don't work nicely with testbot, so let's commit the patch and then run the test again just to confirm it really was that. Looks fine to me otherwise, thanks Steven.