When a container is marked as being in STOPPED state it is possible for another
thread to start it again even though not all cleanup operations for that
container have finished. This is not a problem for most things like sockets and
friends which are unique to the container. It is however a problem for cgroups
which are named after that container in that we cause our cgroup driver to
waste cpu cycles finding a new cgroup name.
Closes #1726.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>