Add Python gRPC backend using WhisperX for speech-to-text with word-level timestamps, forced alignment, and speaker diarization via pyannote-audio when HF_TOKEN is provided. Signed-off-by: eureka928 <meobius123@gmail.com>