Interactive Real-Time Speaker Diarization Correction with Human Feedback