Make alarm slightly less chatty

Tried 2 or 3 hours, but too many emails and worker pool wouldn't have time to stay constrained

 8 hrs is enough time to fully ramp up the worker pool and ramp up jobs on new nodes; only then can we decide that capacity is really constrained in a way that requires changing the scaling policy for that week.

modules/imputation-server/monitoring.tf

@@ -9,8 +9,8 @@ resource "aws_cloudwatch_metric_alarm" "cluster_needs_resources" {
   comparison_operator = "GreaterThanOrEqualToThreshold"
   threshold           = 1
 
-  datapoints_to_alarm = 2
-  evaluation_periods  = 2
+  datapoints_to_alarm = 8
+  evaluation_periods  = 8
 
   actions_enabled = true