TranscodingAPI: Handling Transcding service crash.

Upon receiving a binder died event of the client
due to service crash, we will do the following:
 1) For the job that is running, notify the client
    that the job is failed with error code, so client
    could choose to retry the job or not.
 2) For the jobs that is still pending or paused,
    we will resubmit the job internally once we
    successfully reconnect to the service and
    register a new client.
 3) When trying to connect to the service and register
    a new client. The service may need time to reboot
    or never boot up again. So we will retry for a number
    of times. If we still could not connect, we will notify
    client job failure for the pending and paused jobs.

Bug: 161469320
Bug: 160260102
Test: Run the test with long clip and kill the service.
Change-Id: Idfd8dd0aae60b1dfd6c766552c71f07c60663918