Update batching logic.

1. Attempt to piggyback on network activations.
   Have JobScheduler attempt to delay the start of some connectivity
   jobs until the network is actually active. This should help reduce
   the number of network wakeups and the cumulative network
   idle-to-suspend duration.
2. Include jobs in the ACTIVE bucket in the forced batch set and don't
   force batch if we're too close to a job's deadline. However, still
   continue exempt jobs for apps in the EXEMPTED bucket and jobs for
   apps that are currently active (uidActive).

Bug: 28382445
Bug: 299329948
Bug: 299346198
Test: atest frameworks/base/services/tests/mockingservicestests/src/com/android/server/job
Test: atest frameworks/base/services/tests/servicestests/src/com/android/server/job
Test: atest CtsJobSchedulerTestCases
Change-Id: I58f47c8297b43e72a0b0af16aaf71ace3996166f