Feed HTTP syncing request buffers iterativelly to avoid keeping syncing docs in memory
Modify the SyncExchange used by Soledad so that the documents are iterativelly delivered to the HTTPResponder. This way, we don't need to maintain all fetched docs in memory while also having them represented as strings in the twisted WSGI HTTP response buffer.
- http://bazaar.launchpad.net/~ubuntuone-control-tower/u1db/trunk/view/head:/u1db/remote/http_app.py#L439
- http://bazaar.launchpad.net/~ubuntuone-control-tower/u1db/trunk/view/head:/u1db/sync.py#L250
(from redmine: created on 2014-03-18, closed on 2014-03-20)