Medium: Overhaul the way peers are determined to be sufficiently active

• Make the implementation of crm_is_peer_active() stack-dependant
• Move more heartbeat code out of common files
• Corosync 2.0 and cman based cluster no longer uses process list data from the mcp. They now use cpg membership to determine if a peer process is active
• Differentiate between the proc entries for each communication type instead of re-use crm_proc_ais
• Always create a node if none exists when returning from crm_update_peer()