Resource templates overcommit memory and can throttle CPU #912

The resource presets offered to users can lead to OOM errors under high contention, besides that it's debatable whether CPU limits are a good idea in general. It makes sense to fix all the resource presets like so:

   "nano" (dict 
       "requests" (dict "cpu" "100m" "memory" "128Mi" "ephemeral-storage" "50Mi")
-      "limits" (dict "cpu" "150m" "memory" "192Mi" "ephemeral-storage" "2Gi")
+      "limits" (dict "memory" "128Mi" "ephemeral-storage" "2Gi")
    )

removing the cpu limits and setting the memory limits equal to requests.