gbq has bad performance

[phurst]

As a result of Net::CUPS's getJob method having performance O(number of active jobs + number of completed jobs), gbq is slow. The performance isn't the worst thing in the world, but we could do better; we can get I believe O(1) performance to get the attributes of current jobs if we're willing to speak raw IPP (this is what pycups does), but we can still get somewhat decent performance if we avoid calling getJob once for each active job, or tell it 'hey only look at active jobs' or something.