We should update the CIDB status for builds from INFLIGHT to something new (EXPIRED has been proposed) after the CIDB timeout for the build has been reached.

This is a catchall for builds that are aborted by means we can't track (such as buildbot aborts, buildbot timeouts, spontaneous builder reboot, etc).

Possible implementations discussed:

A) CIDB cleanup cron job is created.
B) master builders run new CIDB cleanup step.
C) MySQL solution (virtual table, etc).