[no ticket] Use higher ASG targets (#3044)

## Context

At 50% memory, the frontend hits that memory cap easily and immediately
gets cycled out. I wasn't really expecting this behavior, but you live
and you learn.

## Changes proposed

Increases the memory target to 99%, so nextjs only spins up new
containers when after containers max out their cache. This won't
meaningfully impact the backend, because the because the backend doesn't
have an infinitely* scaling cache.

\* There's not significant indication that the nextjs cache is
infinitely scaling. It might just scale higher than our current memory
numbers.

I also increased the CPU target to 75%, which may come back to bite me I
dunno. We'll see.

infra/modules/service/autoscaling.tf

@@ -25,7 +25,7 @@ resource "aws_appautoscaling_policy" "ecs_scale_policy_memory" {
     }
     scale_in_cooldown  = 300
     scale_out_cooldown = 5
-    target_value       = 50
+    target_value       = 99
   }
 }
 
@@ -42,6 +42,6 @@ resource "aws_appautoscaling_policy" "ecs_scale_policy_cpu" {
     }
     scale_in_cooldown  = 300
     scale_out_cooldown = 5
-    target_value       = 50
+    target_value       = 75
   }
 }