Make cluster auto-recover

[Hades32]

With these changes (and the update to v0.25) this setup should finally reliably work.

• The reset-peers flag is needed to recover from pod IP changes
• read/write-lag flags are made explicit as, depending on cluster topology, folks will want different values here
• a readiness probe fixes consumers hitting dead nodes and causing even more issues
• the publishNotReadyAddresses setting ensures that pod IPs are still available for clustering

[wille]

Thanks,
I also had to set

          args:
            - --peering-address
            - $(POD_IP)
          env:
            - name: POD_IP
              valueFrom:
                fieldRef:
                  fieldPath: status.podIP

[IT-Guy007]

Also update the readme if approved, the config works for me on Google Cloud.