AWS: kube-down was failing with no instances

The new delete-ASG logic wasn't correct if there were no instances

cluster/aws/util.sh

@@ -1204,6 +1204,7 @@ function kube-down {
                                       Name=tag:KubernetesCluster,Values=${CLUSTER_ID} \
                             --query Reservations[].Instances[].InstanceId)
 
+    if [[ -n "${instance_ids}" ]]; then
       asg_groups=$($AWS_CMD   --output text describe-instances \
                               --query 'Reservations[].Instances[].Tags[?Key==`aws:autoscaling:groupName`].Value[]' \
                               --instance-ids ${instance_ids})
@@ -1218,7 +1219,6 @@ function kube-down {
         fi
       done
 
-    if [[ -n "${instance_ids}" ]]; then
       $AWS_CMD terminate-instances --instance-ids ${instance_ids} > $LOG
       echo "Waiting for instances to be deleted"
       while true; do