End-to-End Multi-Microphone Speaker Extraction Using Relative Transfer Functions