Make Java service discovery Redis Cluster compatible

[zhchxiao123]

What to build

Make Java service registration and discovery work correctly when the SDK is configured for Redis Cluster. Services should be able to register, heartbeat, unregister, refresh cached instance lists, and discover healthy instances through the same public APIs used in standalone mode.

The implementation should use the existing cluster-aware Redis abstraction or an equivalent business-level abstraction instead of borrowing a standalone Redis resource directly. Keep same-service keys aligned with the v2 key schema so the active-instance index and instance-detail hash remain safe under Cluster.

Acceptance criteria

• Registering a service instance succeeds in Redis Cluster mode and makes the instance visible to discovery clients.
• Heartbeats update the active-instance timestamp in Redis Cluster mode without relying on a standalone connection pool.
• Unregistering a service removes both the instance details and active-instance entry in Redis Cluster mode.
• Discovery refresh reads active instance IDs and details in Redis Cluster mode and preserves health-threshold filtering.
• Existing standalone service discovery behavior remains covered by tests.
• Tests cover the Cluster path without direct calls to a standalone-only Redis resource.

Blocked by

• Enable Redis Cluster initialization for Java client #31