Sql must process unstable topology cases and recovery request if it possible

During cluster life cycle this is normal that some nodes are leaving topology and possibly come back at future. Ignite need to process such cases during a continuous stream of sql requests.