Full synchronous replication

The task stands for further enhancement of the semisync replication and addresses scalability and performance. It deals with the following aspects.

• transactions are sent out from binlog group commit module (so grouped) in parallel
  with syncing them to the binary log (when the option is on)
• the sender part adopts a group communication system to synchronize the sender and receivers, possibly optimize network operations (like make use of multicast)
• the total master plus slave set of servers "self-" organizes into a cluster group to
  be notified on membership change (which event can be propagated to a client)
• flexible acknowledgement to require a configurable # of ack:s (NO, 1, majority, weighted quorum).

This task is an umbrella for the set of features highlighted.