This PR adds support for agent shutdown lifecycle states and shutdown_script.

It (currently) relies on terraform provider abilities to gracefully shut down. For instance with the docker provider it can be controlled via:

resource "docker_container" "workspace" {
  destroy_grace_seconds = 30
  stop_timeout          = 30
  stop_signal           = "SIGINT"
}

The agent can handle common signals like SIGHUP, SIGINT, SIGTERM. If need be, more signals could be handled.

Builds upon #5171
Fixes #4677

Note:

• In hindsight, I should probably have called the lifecycle states stopping, stop_timeout, stop_error, stopped, yay or nay to make this change?