CI: Raft consensus testing under failure conditions

[renecannao]

Summary

Orchestrator supports Raft-based HA deployments with multiple nodes forming a consensus cluster. The Raft code (go/raft/) is significant but has zero CI coverage for failure scenarios like leader election, leader failure, and quorum loss.

Proposal

• Set up a 3-node orchestrator Raft cluster in Docker
• Test leader election on startup
• Test leader failure and re-election
• Test that topology operations are only served by the leader
• Test that followers redirect to the leader

Acceptance Criteria

• 3-node Raft cluster starts and elects a leader
• Killing the leader triggers re-election within timeout
• API requests to followers return redirect to leader
• Topology operations succeed on new leader after failover
• Test integrated into CI workflow