fix(telemetry): schedule the trace payload send instead of dropping it

close_trace() called the coroutine TelemetryManager.start_send_task without
awaiting or scheduling it, so every payload closed on that path was silently
dropped and the runtime logged "coroutine ... was never awaited". The only
other caller awaits it correctly, so both call styles stay supported: call the
manager, schedule the returned coroutine on the running loop, and close it with
a debug log when no loop is available.

Regression case added in tests/unit_tests/telemetry/test_trace.py: it fails
before this change (payload never delivered) and passes after.

src/langbot/pkg/telemetry/execution.py

@@ -167,10 +167,28 @@ class ExecutionCounters:
                 return
             payload = self._build_trace_payload(state, reason)
             if payload is not None:
-                self.manager.start_send_task(payload)
+                self._dispatch_payload(payload)
         except Exception:
             return
 
+    def _dispatch_payload(self, payload: dict) -> None:
+        """Hand one built payload to the telemetry manager from this sync context.
+
+        TelemetryManager.start_send_task is a coroutine, so calling it without
+        scheduling dropped every trace closed here. Stand-ins used by tests and
+        manual tools schedule synchronously, hence the coroutine check.
+        """
+        result = self.manager.start_send_task(payload)
+        if not asyncio.iscoroutine(result):
+            return
+        try:
+            asyncio.get_running_loop().create_task(result)
+        except RuntimeError:
+            result.close()
+            logger = getattr(getattr(self.manager, 'ap', None), 'logger', None)
+            if logger is not None:
+                logger.debug('Execution trace payload dropped: no running event loop')
+
     def _trace_emitted(self, state: TraceState) -> bool:
         mode = self.trace_mode()
         if mode == 'off':