test: assist with debugging QEMU startup

When kubeadm fails, the error is followed by lots of diagnostics. That's good
for the CI, but can hide the actual error. Some extra message at the end helps.

When running interactively, pausing the shutdown gives an opportunity to debug
without having to start again with a modified script.

test/setup-kubernetes.sh

@@ -252,6 +252,7 @@ sudo kubeadm init $kubeadm_args $kubeadm_args_init || (
     # Dump some information that might explain the failure.
     sudo systemctl status docker crio containerd kubelet
     sudo journalctl -xe -u docker -u crio -u containerd -u kubelet
+    echo "ERROR: kubeadm init failed, see above."
     exit 1
 )
 mkdir -p $HOME/.kube

test/start-kubernetes.sh

@@ -523,6 +523,12 @@ function cleanup() (
             echo "Docker log for $name:"
             docker logs $name | sed -e "s/^/$name: /"
         fi
+
+        if [ -t 0 ]; then
+            echo "Press RETURN to clean up."
+            read input
+        fi
+
         for vm in $(govm list -f '{{select (filterRegexp . "Name" "^'$(node_filter ${NODES[@]})'$") "Name"}}'); do
             govm remove "$vm"
         done