End-to-End Evaluation for Low-Latency Simultaneous Speech Translation