stale records for short-lived containers

[rade]

When a container dies just before we tell DNS about it, we end up creating a stale record that is never removed. To reproduce, apply

--- a/weave
+++ b/weave
@@ -684,6 +684,8 @@ put_dns_fqdn() {
     FQDN="$2"
     shift 2

+    docker kill $CONTAINER_ID
+
     for ADDR in "$@" ; do
         http_call $DNS_CONTAINER_NAME $DNS_HTTP_PORT PUT /name/$CONTAINER_ID/${
     done

and then run

$ weave launch
$ weave launch-dns 10.2.1.1/24
$ weave run 10.2.0.1/24 -h foo.weave.local -ti ubuntu
$ weave status | grep "foo.weave.local"
51b239d2adc1: foo.weave.local.[10.2.0.1]

That last output is proof of the stale record.

[rade]

As with #819, the suggested fix is to check (using the docker api) whether the container is alive after adding it to the datastructure which is cleaned up by the docker event stream handler.