[dns] failure during launch-dns can lead to long pause

[bboreham]

If the weavedns container fails during start-up, but after the script gets into populate_dns, then it will take about 5 seconds to bail out because it doesn't check to see if the container is still running; it just repeatedly calls curl then sleep.

Also it doesn't check to see if a valid IP was returned from 'docker inspect'.

[rade]

Looks like things have moved on since this was filed...

1. We now wait for the dns container to respond to a status request, before doing anything else. This will catch any issues during dns container startup.
2. Following that, we just invoke 'curl' for each container IP. That invocation has a configured connect timeout of 3 seconds. If it fails the weave script exits with an error code. There is no message though; I shall fix that.