Avoid in-place index upgrades to reduce server downtime.

To avoid what happen in our last sprint review, i.e. a last minute deployment for one story breaking the demo for another, this integrates the postinst script and the indexer to perform index upgrades by writing a completely new index and only when that is finished renaming the existing one and restarting the server so the already running continues to work while the indexer is running. (This would have failed before as the server would automatically reload the newly created and therefore empty index (with a potentially incompatible schema if it were filled to boot).)

Closes #239 (closed)