Simplify Adhoc#start_inactive_instance logic

[wjordan]

When starting a stopped EC2 instance its public IP address changes, because IPv4 addresses are not retained for inactive network interfaces. In order for an adhoc instance (which does not use a CDN or load-balancer) to continue serving public traffic, its DNS entries need to be updated to point to the new IP addresses when the instance is started again.

Previously, the adhoc:start_inactive_instance rake task would call #start on the EC2 instance, then manually update the DNS entries linked from the CloudFormation stack via change_resource_record_sets. This had an issue where it would prevent the DNS records from deleting when the stack is deleted, because the AWS::Route53::RecordSetGroup CloudFormation resources would become out of sync with the actual entries. The issue prevents adhoc stacks from being deleted and re-created due to the left-over DNS entry.

This PR fixes the issue and simplifies the start_inactive_instance implementation by modifying the UserData property of the daemon instance and updating the stack. This triggers a reboot of the EC2 instance (causing it to start), and also triggers the RecordSetGroup resources to be updated to the instance's new values.