Scale down systests

Use cluster size 20 instead of 30 and increase timeout to 75 min from
60. We should try to increase cluster size again once we figure out
the reason for performance regression in systests.

.github/workflows/systest.yml

@@ -130,13 +130,13 @@ jobs:
           go-version-file: "go.mod"
 
       - name: Run tests
-        timeout-minutes: 60
+        timeout-minutes: 75
         env:
           test_id: systest-${{ steps.vars.outputs.sha_short }}
           label: sanity
           storage: premium-rwo=10Gi
           node_selector: cloud.google.com/gke-nodepool=systemtest
-          size: 30
+          size: 20
           bootstrap: 4m
           level: ${{ inputs.log_level }}
           clusters: 4