Currently the task containers will reach a conclusive state (success or
failure), and will promptly stop running. This creates the following
problems:
1. In the case of failure, a user doesn't have the ability to exec into
the container for debugging purposes.
2. In the case a success, the readiness probe for each task container
will begin to fail. This happens because readiness is defined by
the existence of a file; if the container stops running, the file
no longer exists.
This change provides a solution to the 2nd bullet, keeping the container
alive in case of success.
Change-Id: Iae432caada3d2a338493b1534da72957808185d2