Support affinity/anti-affinity policies for VDUs

"VNF clustering feature" is comming to be supported and it has a plan to implement anti-affinity by AZ separation. It seems good approach from a point of view of BCP use case for the services provided to many people.
On the other hand, there is another HA use case that VNF such as vCPE which provides service to individual users want to prepare for a equipment failure (e.g. hypervisor errors.) In such case, the Active and Standby instances have been often placed closely and monitored the other's health each other by some dedicated protocol like VRRP to detect failure and fail-over quickly and automatically. The current VNF clustering feature seems to be difficult applying to such VNFs because active and standby pair are assumed to be placed in the same AZ and their failover doesn't depend on LB.
To complement the case, this feature enables to define affinity/anti-affinity relationship among VDUs in the topology model. It will also help to define affinity of a set of VNF with the VNF clustering feature.