Ml service memory leak

[myungjoo]

Lazily initialize the internal message thread and queue for ml_service_extension to prevent memory growth on repeated new/destroy calls.

Previously, ml_service_new() for extension services would always create an internal message queue and worker thread, even if no actual requests were made. Rapidly calling ml_service_new() and ml_service_destroy() would lead to accumulated thread/allocator cache resources (e.g., from GLib/GObject) that were not immediately released by the OS, causing the RSS to continuously increase and appear as a memory leak. This change ensures the thread and queue are only created on the first actual request.

---

 

[cursor]

Cursor Agent can help with this pull request. Just @cursor in comments and I'll start working on changes in this branch.
_{Learn more about Cursor Agents}