BGP-based implementation for Flexible IPAM

[antoninbas]

If we interpret "Flexible IPAM" as the ability to assign arbitrary IPs to Pods, without looking at the current implementation (which requires bridging mode), then maybe this is something that BGP can actually help achieve (when supported by the underlay) without requiring that the Node's transport interface be moved to the OVS bridge?

Note: that would be a L3 approach to Flexible IPAM, not a L2 approach using VLANs (current approach).

Originally posted by @antoninbas in #6203 (comment)

As pointed out by @tnqn in the above discussion, that would generally require that Nodes be able to learn routes from BGP peers, which was out-of-scope of our initial BGP implementation (#5948). This could also enable additional use cases, e.g. for multicluster routing.