Graceful Restart (GR) and LongLived GR (LLGR) in contrail-control module

In Release 3.2, provide limited support to Graceful Restart (GR) and Long Lived Graceful Restart (LLGR) helper modes to contrail-controller module.

There are two pieces to this. One is when contrail-module itself (gracefully) restarts, then one should be able to avail GR helper functionality provided by its bgp peers. Other is when the peers (bgp and/or xmpp) restart, provide GR helper mode in order to minimize impact to the network. This is achieved using the standard mark and sweep approach to manage the learnt (stale) information.

This feature along with "head-less" mode in the agent should enable existing network traffic to be unaffected when all contrail-control daemons go down. This would be applicable to both east-east traffic between contrail-vrouters and north-south traffic between contrail-vrouters and the SDN Gateways (e.g. MX)

Even during agent restart, flows in the ingress nodes may not be affected as routing-changes are not propagated to ingress agent peers. However, traffic will be affected when agent comes back up, resets the data plane and then reprograms the vrouter in the data plane. This shall be further optimized in future releases