Use infrastructure_provisioning.sh instead of setup_cluster.sh.
The code to save and reuse terraform.tfstate in /data/infrastructure_provisioning/ goes into infrastructure_provisioning.sh. (From there, the good old setup_cluster.sh is called too.)
Rename destroy_cluster.sh to infrastructure_teardown.sh
In etc/longevity.yml, this introduces the saving of terraform.tfstate under /data/infrastructure_provisioning/. However, we still call infrastructure_teardown.sh at the end. The distro used in mongo-longevity doesn't currently specify a teardown hook, so there's nobody there to actually use the data from /data/infrastructure_provisioning/.