Improve visibility of plugin startup problems

[bboreham]

If the plugin detects a fatal error condition and exits, it gets restarted by Docker because of the --restart=always. However this situation is not made very clear to the user.

For example, a bad option will cause an immediate exit:

$ weave launch-plugin --foo
7c4094d284ecec15addf8a2e800cc42a2509db51f4675e5a7479eeae45a03731

No hint here, and worse, weave status gives a normal report:

$ weave status
[...]
        Service: plugin
     DriverName: weave

Compare:

$ weave launch-router --foo
The weave container has died. Consult the container logs for further details.

Note #1814 fixed some related things in the weave script.