Add droplet monitor alert for cloud-init completion

[Tythos]

Is your feature request related to a problem? Please describe.

Currently, cloud-init passed via user_data continues for some time beyond when Terraform considers applying a droplet to be "completed. This is reasonable and expected. However, there are a variety of other actions that may require specific cloud-init actions to have completed (package installation, asset fetching from spaces buckets, etc.). Instead of using a provisioner workaround (which is described by Hashicorp as a "last resort" on https://developer.hashicorp.com/terraform/language/resources/provisioners/syntax and has similar issues with signaling execution completion), a more robust path forward would be to leverage the existing monitor alert system.

Describe the solution you'd like

Adding a 'cloud-init-complete' (or similar) feature would let Terraform procedurally construct such alerts as part of the deployment workflow. Currently, this would result in email or slack notifications, which introduces a manual dependency to the deployment path. Optimally, this could be used by on-system scripts or other resource activities (in a way that would currently be checked by "while..." assertion of ps -e | grep cloud results). So, there are two specific modifications I think are relevant:

1. Addition of a v1/insights/cloud_init_complete to the monitor alert resource types (listed on https://registry.terraform.io/providers/digitalocean/digitalocean/latest/docs/resources/monitor_alert ); under the hood, this can be triggered with a simple check against ps -e | grep cloud results.

2. An additional (third) alerts block option for appropriate triggers--the most straightforward and simple example I can think of is a path to a shell script to invoke, as such a script could be configured/installed as part of the cloud-init specification. (Why couldn't this just be called from the cloud-init specification itself? Good question. I think such scripts wouldn't focus on continued in-environment execution as much as they would be used to propagate actions to/from other systems. For example, null_resource checks could be used by the Terraform executor to trigger subsequent resource applications.)

Describe alternatives you've considered

Alternatives include:

• Adding a poll ("while...") loop in an external script

• Triggering an explicit alert from the end of the cloud-init commands themselves

• Forgoing cloud-init / user_data entirely and forwarding system configuration responsibilities to Ansible, Chef, Puppet, etc. (though setup of such tools would still be needed)

Additional context

Related to #280 , as such alerts could be used to explicitly trigger other downstream dependencies that would otherwise require a "last resort" provisioner. But more specifically, this is intended to expose downstream dependencies for more automated deployment workflows. Common triggers for post-cloudinit actions include following up with Ansible playbooks or other configuration management assertions. More sophisticated triggers could include Terraform callbacks via null_resource checks.

[bartekpacia]

I'd love to see this! :)