production/nixguix: Let loader pods run

It's been sent a warm up shutdown while running. Those tasks are long running. This
seems to prevent the task from finishing properly. Which then cascades to no longer been
scheduled later on (since it could not sent the final status once it's done).

Refs. swh/infra/sysadm-environment#5213

swh/values/production/swh.yaml

@@ -656,7 +656,9 @@ loaders:
           - rpm
           - diff
       autoScaling:
+        stopWhenNoActivity: false
         maxReplicaCount: 2
+        minReplicaCount: 2
     npm:
       requestedMemory: 256Mi
       requestedCpu: 200m