CI: fix: stop the deploy prune racing the rollout, and report why it failed
Two problems, both in the ordering of the remote deploy script.
**The prune ran during the rollout.** docker container prune -f and
docker image prune sat immediately after docker stack deploy, so they fired
while the rolling update was still in flight. That is the worst available
moment: the prune races Swarm's own containers, and any task that has already
exited loses its container -- and therefore its logs -- before an...
#644
| Job | Run time |
|---|---|
| 5m 37s | |
| 5m 37s |