Waiting queue doesn't respect queue processing timeouts.

With the current logic, waiting queue will never actually time its workers out per the processing time limit; this is because the inner while() loop will just keep on repeating. Attached patch fixes it. It now checks after a claim has returned an job to ensure that we're still within the time limit. If we are, it processes the job; if not, it releases the job back to the queue, watchdogs and breaks out of the outer while.