Unified End-to-End Speech Recognition and Endpointing for Fast and Efficient Speech Systems