Implement logic to handle split-brain scenario

At the start of the agent, this change takes into account:

• the presence of CEO's static pod etcd
• a "force_new_cluster" CIB attribute (with lifetime reboot)
• agent status on the peer node (e.g. active, starting, not starting), obtained enabling OCF_RESKEY_CRM_meta_notify
• etcd cluster-id, local and on the peer node
• etcd revision, locally and on the peer node

To decide if the agent should:

• start normally expecting a cluster of 2 nodes (i.e. during CEO handover)
• force a new cluster to recover from a etcd quorum failure
• join and existing cluster-of-1 (and so clear its local etcd data first)