fix: absorb late socket errors after cleanup

[alexx-ftw]

Summary

• keep a permanent socket error listener in IntercomClient.connect() so late read errors after cleanup/destroy cannot crash the Pi process
• add a regression test that simulates an initial ECONNRESET, verifies cleanup leaves one guard listener, and confirms a later ECONNRESET does not throw
• include the new client regression test in npm test

Why

A real Pi session crashed on Node.js v24.14.0 with:

node:events:486
      throw er; // Unhandled 'error' event
Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:216:20)
Emitted 'error' event on Socket instance at:
    at emitErrorNT (node:internal/streams/destroy:170:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)

cleanupSocketListeners() removes the normal socket error handler during disconnect/error cleanup. If Node emits a late socket error while/after socket.destroy() completes, no listener remains and the process crashes. The permanent guard only absorbs errors that would otherwise be unhandled; normal error handling still runs through the existing handlers before cleanup.

Verification

• npm install --ignore-scripts
• npm test → 37 passing tests
• independent review: no Critical/Important/Minor findings