Improve monitoring for secondary priority metrics

[ryanjjung]

Add monitors for metrics not covered by issue #57 but which might occasionally reveal infra problems.

• NLB unhealthy host count (related to RDS module)
• ECS cluster/service alarms pertaining to extended metrics provided by Container Insights:
  • Used vs. reserved CPU/RAM
  • Scale metrics
• SNS topic failed messages (currently used for email notification on alarms, making this a little redundant — if SNS can't notify, we won't be notified that it can't notify — but this will have other uses down the line).
• RDS instances metrics (not used by Send, but I believe used by Appointment), such as:
  • Replication checkpoint lag
  • CPU credit balance
  • CPU/RAM utilization, freeable memory
  • Disk queue depth (to detect disk I/O problems), I/O latency
  • Various network metrics
  • Swap usage
• EC2 instance metrics
  • CPU/RAM utilization
  • EBS volume I/O
  • Network I/O
  • Status check failures