Address persistent GCE instance

[glasnt]

postdeployment.tf uses a compute engine instance's metadata startup script to invoke the Cloud Run jobs created in jobs.tf.

This instance currently keeps running throughout the life of the deployment, and has no purpose. It would be optimal to shutdown this instance after use to entire a lower ongoing cost.

Additionally, metadata startup scripts only fire on instance boot. Re-applying TF where an instance is running won't re-fire this script.

The client job can be re-run without issue, but the setup job isn't currently idempotent, but can be adjusted.

It would be optimal to ensure re-applying TF starts this instance up again, ensuring jobs are re-run to reassert state.

[glasnt]

On postdeployment.tf:

resource "google_compute_instance" "initialize" {
...
   allow_stopping_for_update = true
+  desired_status = "RUNNING"
...
-# shutdown -h now
+shutdown -h now

[glasnt]

Was occasionally getting "cannot assign requested address" errors with this, but they appear to be unrelated, and Emblem also gets them, specifically in Cloud Shell GoogleCloudPlatform/emblem#712

Upstream: hashicorp/terraform-provider-google#6782