BFD based health detection for BGPaaS

Provide BFD support for BGPaaS sessions

Periodically send BFD packets to the currently active BGPaaS session. Typically, there would be many VMI links over which BGPaaS session would come up. But at a given time, only one session is expected to be active (on which BGPaaS runs and exchanges routing information with contrail-control). However, BFD server should be able to maintain BFD sessions on the VMI. (i.e, multiple BFD sessions could existing at some time, with the same source ip, destination ip, within a tenant vrf). VMI should be used to uniquely identify/map BFD sessions correctly

When BFD session goes down, agent effectively brings down the VMI by withdrawing all routes locally originated over that VMI. This potentially results in BGP NextHop for all the routes learned over that BGPaaS session as unresolved and hence unusable. That would automatically result in control-node deleting/withdrawing the BGPaaS routes learned from session (over which BFD had failed). During such a detection of failure, agent can stop forwarding all bgpaas packets to the control-node. This would cause bgp hold timer to expire, and then the bgp session itself would go down automatically as well.

Provide configuration support to enable/disable BFD on a given BGPaaS configuration object.

Reference: https://blueprints.launchpad.net/juniperopenstack/+spec/bfd-over-vmis