Operator should set a recovery type when recovering a Couchbase Server pod

When upgrading a pod via graceful-failover and delta-recovery, the Operator currently sets a recovery type of "delta" if the server node inside the pod supports delta and otherwise the recovery type is left undefined. This is a problem as when the rebalance is run all nodes in inactiveFailed are automatically ejected from the cluster if the recovery type is undefined.

When recovering, the Couchbase Server node must have a recovery type set. It can be delta if the server node supports it; else it must be full.