fix: dispose RPC session and ping timer so CI exits after reporting

[veksen]

Summary

• The analyzer's CI job hung after "Generating report (GitHub)" because the WebSocket RPC stub and the 30s ping `setInterval` opened by `ApiClient.connect` kept the Node event loop alive until GitHub's runner timeout fired.
• `ApiClient.connect` now returns `{ api, dispose }`; `dispose()` clears the ping timer and disposes the (authenticated + unauthenticated) capnweb stubs via `[Symbol.dispose]`.
• `runInCI` wraps the post-connect flow in `try { … } finally { disposeApi(); }` so cleanup runs even if reporting throws; `connectWithReconnect` also calls `dispose()` on `onRpcBroken` to avoid leaking timers across reconnects.
• `runOutsideCI` (Docker server path) is untouched and still uses its own SIGTERM/SIGINT shutdown.

Test plan

• Local repro: run `node --import tsx src/main.ts` with the CI env vars and confirm the process exits within a few seconds of "Generating report (GitHub)" instead of hanging.
• Site CI: trigger the `analyzer / Analyzer` job against this branch and confirm it completes in well under the runner timeout.

[github-actions]

Query Doctor Analysis

View full run details

2 queries analyzed

2 pre-existing issues

• SELECT "guests"."id", "guests"."session_id", "guests"."username", "guests"."avatar_path", "guests"."color", "guests"."side", "guests"."audio_recording_path", "guests"."audio_recording_public", "gue...
  index assets(event_id, inserted_at desc)
  cost 31,003,449 → 1,498 (100% reduction)
• SELECT * FROM guest_ip_addresses WHERE ip_address = '127.0.0.1';
  index guest_ip_addresses(ip_address)
  cost 154,402 → 8 (100% reduction)

_{Using assumed statistics (10000000 rows/table). For better results, sync production stats.}