Projection rebuilding via Space.jobQueue

[rhyslbw]

The Space.jobQueue solution will include a worker for internal background tasks such as this, which should be active by default, but optional to allow these tasks to be moved off the main process. We need to coordinate the rebuild with all application instances since a critical part of the rebuild is holding new events published during the rebuild in memory, then processing them once done. All instances will need to put the nominated projections into this state, then only one instance access the repository and replay the events.

The preparation makes sense to be done either via the distributed commit store or the proposed space:message-queue, and the actual task of replaying the events should be pushed into the job queue so only one worker completes it. If the task times out or fails to complete, it will be failed or will report back as a failure, which then makes it available for another attempt. The complexity here will be knowing when the replay task can proceed, so we may need to consider the discussed option of using meteorhacks:cluster for deployment awareness.
Ensure only one application instance performs the rebuild, but others hold events until complete