Update UnhealthyHost Alarm

[srtalbot]

• if one host is down, send the error message to slack.
• If both hosts are down, send an alarm to opsgenie
• investigate why we didn't get an error message for the high memory usage that hit our cap.

[patheard]

The reason we didn't get an alarm for high memory use is because of how our alarm is configured:
https://github.com/cds-snc/forms-terraform/blob/27de887afba0ce8d80a8bceb3fe22048795781c2/aws/alarms/cloudwatch_app.tf#L26-L30

It is setup to trigger on Average memory use above 50%, and only if that average use is sustained over 4 minutes:

2 evaluation periods * 120 second period

Looking at the period we had expected this alarm to trigger, Average memory use didn't get about 40%, but Maximum memory use hit 100%:

I'd recommend we switch the alarm to Maximum memory use to catch this in the future:

[patheard]

Alarms have all been updated.