CLI command for getting memberlist state

[tnqn]

Describe the problem/challenge you have

The Egress feature and ServiceExternalIP feature rely on memberlist cluster to know the state of Nodes, with which fail-over could be triggered when a Node is gone. Members in the memberlist cluster use a gossip protocol to communicate. Normally all members should have consistent information about the cluster eventually. But there could be implementation bugs or networking problems leading to inconsistency. For example, #4491 fixed a bug that dead Nodes didn't rejoin the cluster if network downtime lasts more than 15 seconds. It also happened several times that the ports required by gossip protocol, TCP and UDP 10351, were not allowed by firewall.

To troubleshoot such issues effeciently, we need to know the state of memberlist cluster of each antrea-agent. A CLI command below would be helpful:

$ antctl get memberlist
NODE      IP            STATUS
worker1   172.16.0.11   Alive
worker2   172.16.0.12   Alive
worker3   172.16.0.13   Dead

With the tool, we could know every antrea-agent's memberlist cluster state and find inconsitency when a brain-split happens.

To simplify troubleshooting, we could add the information to supportbundle as well, meaning antctl supportbundle and the supportbundleCollection API would collect it autotically.