[20160819184515] Cluster Health and Cron Improvements

#1096 fixes a bug that would inadvertently terminate good instances in a clusters with more than 100 instances. Thanks @beedub testing and improving how Convox works at large scale.

Cron gets simpler and more secure in #1099 thanks to @ddollar. The periodic Lambda function that runs your cron jobs now has a more explicit IAM role and uses the ECS RunTask API.