enhance: Implement Data-Path Health Check API for end-to-end log flow validation

[tinswzy]

Description
Currently, Woodpecker (in Service mode) relies on a simple "Ping-Pong" network check for liveness probes. However, this only confirms the process is running and the network is reachable; it does not guarantee that the entire data path (Log read/write) is functional.
We need a more robust Health Check API that validates whether logs can actually be written to and read from the underlying storage/buffer, covering all deployment modes.

Proposed Logic
The new health check should be based on real-time log activity rather than injecting dummy "heartbeat" topics.

1. Log-Level Health

• Active Logs: If a log has incoming write attempts:
  • Success within the last 10 minutes = Healthy.
  • Stalled/Blocked for > 10 minutes = Unhealthy.
• Idle Logs: If no data has been written to a specific log for a while, it should persist its last known health state (Stay Healthy if it was Healthy).

2. Global Service Health

• Partial Success: If at least one log is successfully reading/writing, the global state is Healthy.
• Global Stall: If multiple logs have pending writes but all have been stalled for > 10 minutes, the global state is Unhealthy.
• Idle System (Cold Start/No Traffic): If there is no log activity across the entire system:
  • Fallback to a storage backend check (e.g., HeadBucket or equivalent metadata check for Object Storage).
  • If the storage backend is reachable/writable, the global state is Healthy.