chore: persist active-first peers with debounced writer

[s-h-ubham]

Title

Persist active-first peers to last_peers.json with debounced writer; remove lastKnownPeers

---

Summary

This PR reworks peer persistence to make last_peers.json always reflect the most useful peers for fast bootstrap and robust recovery.
It removes the legacy lastKnownPeers logic and replaces it with a unified, debounced persistence mechanism.

---

Key Changes

Active-first persistence

• last_peers.json now lists:

  1. Active peers first — gathered from Peers() (currently connected peers).
  2. Recently seen peers — collected from the peerstore (seenPeers), acting as fallbacks.

• Entries are deduplicated and capped (e.g., at 200) to prevent file bloat.

Debounced runtime writes

• Peer connect/disconnect events trigger updates to seenPeers and schedule a debounced write, minimizing disk churn.

Graceful shutdown handling

• Introduced an isShuttingDown flag:

  • Close() performs a final synchronous persist.
  • Further background writes are prevented after shutdown begins.

Code cleanup

• Completely removed lastKnownPeers (fields, helpers, locks, and references).
• Applied minor linter and formatting fixes (wsl, nlreturn, gofmt).

---

Issue Addressed

Previously, last_peers.json could be overwritten with stale data during shutdown, leading to:

• Slow reconnections.

• Dialing of dead or outdated peers.

This update ensures last_peers.json always reflects a current, prioritized, and bounded set of peers, improving reconnection reliability and startup speed.

---

Impact

Behavior

• Faster reconnection after restarts — active peers are dialed first.
• Fallback safety retained via recently seen peers.

Compatibility

• No configuration changes required.
• File format remains the same ([]string of multiaddrs).

Performance

• Minimal overhead thanks to debounced async writes and bounded peer list size.

---

Testing & Validation

• Updated unit tests to validate internal peer tracking (numPeers() instead of Peers() where relevant).

• Manual validation steps:

  1. Start a small cluster, connect/disconnect peers.

  2. Confirm last_peers.json ordering: active peers first, inactive peers next.

  3. Graceful shutdown logs:

     Writing last_peers.json (shutdown)
     

     and shows active-first ordering.

  4. Restart nodes with only last_peers.json: observe dialing prioritizes live peers.

---