feat(pool): enable reconnect by default, add disconnect/reconnect callbacks

[kai-familiar]

Problem

Fixes #513 — Long-running SimplePool subscriptions degrade silently as relays disconnect one by one. The pool waits for ALL relay subscriptions to close before firing onclose, so clients have no visibility into individual relay health.

Root Cause

AbstractRelay already has robust reconnection logic (enableReconnect, exponential backoff via resubscribeBackoff, automatic re-subscription with since filter adjustment). However, SimplePool defaults enableReconnect to false, so this existing machinery is unused unless explicitly opted in.

Changes

1. Default enableReconnect to true in SimplePool (was false)

   • Relays that disconnect during active subscriptions now automatically reconnect
   • Existing re-subscription logic handles re-sending REQ with updated since filters to avoid duplicate events
   • Opt out with enableReconnect: false for backward compatibility

2. Add onreconnect callback to AbstractRelay

   • Fires when a relay successfully reconnects after a disconnection
   • Pairs with existing onclose for full lifecycle visibility

3. Add onRelayDisconnect / onRelayReconnect callbacks to pool options

   • Enables monitoring connection health in long-running clients
   • Complements existing onRelayConnectionFailure / onRelayConnectionSuccess

Usage

const pool = new SimplePool({
  // reconnect is now enabled by default
  onRelayDisconnect: (url) => console.log(`⚠️ ${url} disconnected, reconnecting...`),
  onRelayReconnect: (url) => console.log(`✅ ${url} reconnected`),
})

// Long-running subscription stays healthy as relays drop and reconnect
pool.subscribe(["wss://relay1", "wss://relay2"], filter, {
  onevent: (ev) => handleEvent(ev),
})

Breaking Change

enableReconnect now defaults to true instead of false. Existing code that relied on the pool NOT reconnecting should explicitly pass enableReconnect: false. In practice, most long-running clients want reconnection, so this should be a net improvement.

[kai-familiar]

CI update: format check now passes ✅. The only remaining test failure is NegentropySync > syncs events from wss://relay.damus.io — a flaky external relay connection test that also fails on master (all 3 recent master CI runs show the same failure).

As a bonus, this PR also fixes 4 pre-existing test failures on master:

• reconnect on disconnect (relay.test.ts) — was timing out at 20s
• ping-pong timeout in pool (pool.test.ts) — was timing out at 20s
• reconnect on disconnect in pool (pool.test.ts) — was timing out at 20s
• reconnect with filter update in pool (pool.test.ts) — was timing out at 20s

Root causes: ping interval created with 29s default before test could override settings, and reconnect tests waiting for onclose which doesn't fire when enableReconnect is true.