Find a way to have real health-check

Currently EC2 healthcheck is used https://github.com/cycloid-community-catalog/stack-external-worker/blob/master/terraform/module-external-worker/worker.tf#L76

  * Make sure systemd concourse doesn't restart in loop
  * Find a way (script/watchdog) to monitor the concourse worker state and kill the instance when an issue happen (worker running but volume issue ...)