Ensure Read/Write Safety During Recovery

[vazois]

This PR contains the following enchantments:

• [x] Added validity checks on nodeId and workerId at IsLocalExpensive.
• [x] Add nullptr check to prevent object nullptr ref for concurrent failover and recovery operations.
• [x] Propagate RESP errors back to the client at CLUSTER REPLICATE.
• [x] Ensure SuspendRecovery happens at the end of BeginRecover.
• [x] Prevent start of recovery if there is an ongoing failover.
• [x] Refactor and simplify SingleKeyVerify validation.
• [x] Add recovery check at SingleKeyReadWriteVerify to prevent writes while replica is taking over.