fix(ai): standardize lock handling in _ai_invoke_callback

Take the global lock mutex before invoking all callbacks (success
and error) to ensure thread-safety. Previously some paths took the
lock and others did not, creating asymmetric behavior that could
lead to data races when callbacks access UI state.

src/ai/ai_client.c

@@ -111,8 +111,10 @@ _ai_invoke_callback(ai_callback_data_t* cb_data)
 {
     log_debug("[AI-CALLBACK] _ai_invoke_callback: is_error=%d, response_cb=%p, error_cb=%p",
               cb_data->is_error, (void*)cb_data->response_cb, (void*)cb_data->error_cb);
-    /* profanity uses ncurses, not GLib main loop, so call directly */
+    /* Take the global lock to ensure thread-safety when invoking callbacks that may access UI state. */
+    pthread_mutex_lock(&lock);
     _ai_callback_invoke(cb_data);
+    pthread_mutex_unlock(&lock);
 }
 
 /* ========================================================================