Enforce GCE lifespan during instantiation

FPF used to have a GCE instance culler tool:
 https://github.com/freedomofpress/container-gce-instance-culler
which was packed into a container and run as a cronjob ever other hour
on Codefresh. Instead of trying to maintain said tool and Codefresh
pipeline, define a lifespan during VM creation that GCP will enforce.

devops/gce-nested/gce-start.sh

@@ -43,7 +43,9 @@ function create_sd_ci_gce_instance() {
           --subnet ci-subnet \
           --boot-disk-type=pd-ssd \
           --machine-type="${GCLOUD_MACHINE_TYPE}" \
-          --metadata "ssh-keys=${SSH_USER_NAME}:$(cat $SSH_PUBKEY)"
+          --metadata "ssh-keys=${SSH_USER_NAME}:$(cat $SSH_PUBKEY)" \
+          --instance-termination-action=DELETE \
+          --max-run-duration=3h
 
       # Give box a few more seconds for SSH to become available
       echo "Sleeping for 20s to wait for SSH to become available"