Replication Checkpoint Fix

[vazois]

This PR contains the following changes:

• Separate replication-id used for signing checkpoints vs the replication-id being recovered from an existing checkpoint.
• Set replication-id of ReplicationLogCheckpointManager for signing checkpoints only at initialization or during failover.
• Introduce checks to ensure validity of metadata when send a checkpoint to an attaching.
• Refactor replication info to avoid using explicit count calculations.
• Simplify constructors for ClusterManager, ReplicationManager and FailoverManager.
• Push more fixed error message to CmdStrings.
• Refactor cluster tests to use common backoff for polling the server.
• Misc code cleanup and simplication