Cluster connection failover when cluster topology changes

[mp911de]

lettuce obtains its cluster view once upon starting a connection. There is no dynamic update but updates to the cluster view are on demand possible. Connections which get disconnected in clustered operations keep their queued and buffered commands until the whole cluster connection is shut down and these commands are lots from the execution perspective.

Expected behavior:

• Possibility to obtain dynamic topology changes
• Update the cluster view while clustered operations
• Failover of queued and buffered commands according to command reliability docs

[mp911de]

Topology updates can be activated using ClusterClientOptions by setting refreshClusterView to true

The refresh happens in the background and a validation/activation is performed every time the partitions are loaded using RedisClusterClient#loadPartitions. The method is triggered on the first connection and can be called directly

[mp911de]

Done.