Directory nodes in local index not properly updated on remote deletes

[galfert]

We stumbled over a problem when doing bulk deletes of multiple objects in several directories.

On a second client, there were change events coming in from all the deletes, and the objects themselves were all deleted from the nodes DB in indexedDB. But some of the directory nodes still contained some of the deleted objects in common.itemsMap.

[raucao]

I wasn't able to reproduce this issue yet (because I didn't try with multiple directories), but I did find a different issue with local folders becoming inconsistent after incoming deletes when local changes were present: 098962d

The fix for the other issue was to queue additional sync tasks for paths not appearing in remote itemsMaps, but which are present in local ones, which will then cause sync to see and handle the 404 as a definitive delete (and cleaning up the local itemsMap due to another fix in the mentioned commit). I'm guessing that this should also solve the issue reported here.

However, I'll leave this one open until confirmed.