enos: poweroff and terminate instances when shutting them down (#28316)

Previously our `shutdown_nodes` modules would halt the machine. While
this is useful for simulating a failure it makes cleaning up the halted
machines very slow in AWS.

Instead, we now poweroff the machines and utilize EC2's instance
poweroff handling to immediately terminate the instances.

I've test both scenarios locally utilizing the change and both still
work as expected. I also timed before and after and this change saves 5
MINUTES in total runtime (~40%) for the PR replication scenario. I assume
it yields similar results for autopilot.

Signed-off-by: Ryan Cragun <[email protected]>

enos/modules/shutdown_multiple_nodes/main.tf

@@ -19,7 +19,7 @@ variable "old_hosts" {
 
 resource "enos_remote_exec" "shutdown_multiple_nodes" {
   for_each = var.old_hosts
-  inline   = ["sudo shutdown -H --no-wall; exit 0"]
+  inline   = ["sudo shutdown -P --no-wall; exit 0"]
 
   transport = {
     ssh = {

enos/modules/shutdown_node/main.tf

@@ -19,7 +19,7 @@ variable "host" {
 }
 
 resource "enos_remote_exec" "shutdown_node" {
-  inline = ["sudo shutdown -H --no-wall; exit 0"]
+  inline = ["sudo shutdown -P --no-wall; exit 0"]
 
   transport = {
     ssh = {

enos/modules/target_ec2_instances/main.tf

@@ -186,12 +186,15 @@ resource "aws_security_group" "target" {
 resource "aws_instance" "targets" {
   for_each = local.instances
 
-  ami                    = var.ami_id
-  iam_instance_profile   = aws_iam_instance_profile.target.name
-  instance_type          = local.instance_type
-  key_name               = var.ssh_keypair
-  subnet_id              = data.aws_subnets.vpc.ids[tonumber(each.key) % length(data.aws_subnets.vpc.ids)]
-  vpc_security_group_ids = [aws_security_group.target.id]
+  ami                  = var.ami_id
+  iam_instance_profile = aws_iam_instance_profile.target.name
+  // Some scenarios (autopilot, pr_replication) shutdown instances to simulate failure. In those
+  // cases we should terminate the instance entirely rather than get stuck in stopped limbo.
+  instance_initiated_shutdown_behavior = "terminate"
+  instance_type                        = local.instance_type
+  key_name                             = var.ssh_keypair
+  subnet_id                            = data.aws_subnets.vpc.ids[tonumber(each.key) % length(data.aws_subnets.vpc.ids)]
+  vpc_security_group_ids               = [aws_security_group.target.id]
 
   tags = merge(
     var.common_tags,