Implement response handling and speech tracking in CallEnd functionality

- Add begin_response and finish_after_current_speech methods to CallEndCoordinator for better management of speech events.
- Update PromptBrain to utilize new methods, ensuring proper handling of generated closing speech and tool-only calls.
- Enhance tests to verify the correct behavior of speech tracking and response handling in various scenarios, including waiting for audio to finish before ending calls.
- Introduce a new test suite for CallEndCoordinator to validate the interaction with speech frames.

backend/services/pipecat/call_lifecycle.py

@@ -17,6 +17,7 @@ class CallEndCoordinator:
         self._ending = False
         self._armed = False
         self._speaking = False
+        self._response_speech_started = False
         self._finished = False
         self._reason = "completed"
 
@@ -28,7 +29,22 @@ class CallEndCoordinator:
         self._ending = True
         self._reason = reason or "completed"
 
+    def begin_response(self) -> None:
+        """Start tracking speech produced by one LLM response."""
+        self._response_speech_started = False
+
     def arm_after_speech(self) -> None:
+        """Wait for the next observed bot speech to finish."""
+        self._armed = True
+
+    async def finish_after_current_speech(self, *, has_text: bool) -> None:
+        """Finish now if speech is absent/done, otherwise wait for its stop."""
+        if not has_text:
+            await self.finish()
+            return
+        if self._response_speech_started and not self._speaking:
+            await self.finish()
+            return
         self._armed = True
 
     async def finish(self) -> None:
@@ -38,15 +54,14 @@ class CallEndCoordinator:
         await self._queue_end(self._reason)
 
     async def observe(self, frame) -> None:
-        if isinstance(frame, BotStartedSpeakingFrame) and self._armed:
+        if isinstance(frame, BotStartedSpeakingFrame):
             self._speaking = True
-        elif (
-            isinstance(frame, BotStoppedSpeakingFrame)
-            and self._armed
-            and self._speaking
-        ):
-            logger.info("结束语播报完毕,挂断通话")
-            await self.finish()
+            self._response_speech_started = True
+        elif isinstance(frame, BotStoppedSpeakingFrame) and self._speaking:
+            self._speaking = False
+            if self._armed:
+                logger.info("结束语播报完毕,挂断通话")
+                await self.finish()
 
 
 class EndCallAfterSpeechProcessor(FrameProcessor):