Handle deployment thresholds

[ejholmes]

ECS recently added support for setting deployment thresholds for an ECS service:

https://aws.amazon.com/blogs/compute/amazon-ecs-launches-new-deployment-capabilities-cloudwatch-metrics-singapore-and-frankfurt-regions/

This can be a breaking/annoying change for previous Empire deployments; the old behavior of ECS was to start 1 new task, then kill 1 old one, repeating this process until the service consists of all new tasks. ECS now defaults to starting all new tasks (whatever # the desired count is) then killing the old tasks. It seems like the new default can cause problems if you don't have enough capacity in the ASG.