cloudbank: Explicitly specify what nodepools users should go to

Otherwise, they still end up on the old 'user' nodepool, and it
can not be decomissioned. With this, they will end up on the
equivalent 'new' nodepool, and then I can get rid of the old one.

I have also had to adjust the CPU limit, as at some point I suppose
we switched from 8 CPU nodes to 4 but did not adjust this.

Author: yuvipanda

config/clusters/cloudbank/common.values.yaml

@@ -9,12 +9,15 @@ nfs:
 jupyterhub:
   singleuser:
     cpu:
-      # Each node has about 8 CPUs total, and if we limit users to no more than
-      # 4, no single user can take down a full node by themselves. We have to
+      # Each node has about 4 CPUs total, and if we limit users to no more than
+      # 2, no single user can take down a full node by themselves. We have to
       # set the guarantee to *something*, otherwise it is set to be equal
       # to the limit! We don't explicitly set a guarantee, because there is
       # a guarantee of 0.05 set in basehub/values.yaml
-      limit: 4
+      limit: 2
     image:
       name: quay.io/2i2c/cloudbank-data8-image
       tag: d2746e55a4ee
+    nodeSelector:
+      # Put everything on the most appropriate nodepool for these users
+      cloud.google.com/gke-nodepool: nb-n2-highmem-4