Fix issue #28035090: Disallow abuse of JobScheduler

We now keep track of how long each app has been running a job
for, in 30 minute batches.  If it is running jobs frequently,
we will bump down the priority its jobs run at to allow other
jobs to run before it.

Currently we count both pending and active as the job running,
which means that an app that has jobs waiting in the pending
queue will count against its abuse prevention.  This could
allow starvation -- if we bump down the priority of an app's
jobs and the system is so busy continually that they sit
in the pending queue a lot -- it could never recover.  But I
think that is okay...  if we are really in a state where we
are continually running as many jobs as possible, we probably
have other larger issues.

Change-Id: I838aa4b5840e91df49a1e17b53188d6e4a66a6d1