Feat: add stream inference (Inference Overview of Voice Cloning and Inference Overview of Controlled Generation)

[weedge]

feat:

• add stream inference (Inference Overview of Voice Cloning and Inference Overview of Controlled Generation)

# Inference Overview of Controlled Generation
PYTHONPATH=./ python cli/inference_stream.py \
    --text "身临其境，换新体验。塑造开源语音合成新范式，让智能语音更自然。" \
    --save_dir "example/results" \
    --model_dir ../../pretrained_models/SparkTTS-0.5B \
    --gender female --pitch  moderate --speed high

# Inference Overview of Voice Cloning
# default use static batch is ok
PYTHONPATH=./ python cli/inference_stream.py \
    --text "身临其境，换新体验。塑造开源语音合成新范式，让智能语音更自然。" \
    --save_dir "example/results" \
    --model_dir ../../pretrained_models/SparkTTS-0.5B \
    --prompt_text "吃燕窝就选燕之屋，本节目由26年专注高品质燕窝的燕之屋冠名播出。豆奶牛奶换着喝，营养更均衡，本节目由豆本豆豆奶特约播出。" \
    --prompt_speech_path "example/prompt_audio.wav"

[whaozl]

Hello, could you clarify whether this 'merge' refers to real-time streaming (processed and played simultaneously, rather than being played after full synthesis)?"
@weedge