Figure 3 for Long-Running Speech Recognizer:An End-to-End Multi-Task Learning Framework for Online ASR and VAD