With OVN interconnect, DPU setups are currently broken. 

With wizhao@redhat.com and bnemeth@redhat.com we decided on the way forward:

•  ovnkube-node-dpu-host: it will not connect to sbdb to check the zone name, but will simply read its input zone and will annotate the node with it (as it does already); the purpose is to remove the need for a connection between ovnkube-node-dpu-host on the x86 nodes to SBDB on the DPUs, only for the purpose of retrieving the zone name (ovn-controller doesn't run in ovnkube-node-dpu-host pod, so we'd be effectively cutting off ovnkube-node-dpu-host from sbdb)

• DPU network operator will need a new YAML in order to deploy the full-stack ovnkube-node for the DPU nodes, much in the same way as CNO deploys multizone ovnkube-node
• ovnk node in DPU mode: no input --zone parameter needed: it'll take the zone from the node annotation and validate it against the zone in its sbdb