feat: add local llm -> transformers core llm to support local llm inference, next step for local vision llm

[weedge]

feat:

• add manual transformers core llm
• add pipeline transformers core llm
• add test_transformers for llm_transformers_manual and llm_transformers_pipeline
• add llm_transformers_manual,llm_transformers_pipeline to run local terminal/remote chat

  
  TQDM_DISABLE=True TOKENIZERS_PARALLELISM=false \
  AUDIO_IN_STREAM_TAG=pyaudio_in_stream \
  AUDIO_OUT_STREAM_TAG=pyaudio_out_stream \
  VAD_DETECTOR_TAG=webrtc_silero_vad \
  RECORDER_TAG=vad_recorder \
  ASR_TAG=sense_voice_asr \
  ASR_LANG=zn \
  ASR_MODEL_NAME_OR_PATH=./models/FunAudioLLM/SenseVoiceSmall \
  LLM_TAG=llm_transformers_manual \
  LLM_MODEL_NAME_OR_PATH=./models/Qwen/Qwen2-0.5B-Instruct \
  TTS_TAG=tts_edge \
  python -m src.cmd.local-terminal-chat.generate_audio2audio > ./log/std_out.log

TQDM_DISABLE=True TOKENIZERS_PARALLELISM=false \ AUDIO_IN_STREAM_TAG=pyaudio_in_stream \ AUDIO_OUT_STREAM_TAG=pyaudio_out_stream \ VAD_DETECTOR_TAG=webrtc_silero_vad \ RECORDER_TAG=vad_recorder \ ASR_TAG=sense_voice_asr \ ASR_LANG=zn \ ASR_MODEL_NAME_OR_PATH=./models/FunAudioLLM/SenseVoiceSmall \ LLM_TAG=llm_transformers_pipeline \ LLM_MODEL_NAME_OR_PATH=./models/Qwen/Qwen2-0.5B-Instruct \ TTS_TAG=tts_edge \ python -m src.cmd.local-terminal-chat.generate_audio2audio > ./log/std_out.log

fix:
- log support LOG_LEVEL env to set log level
- check llm stream text have special char for sentences (simple process with re, maybe use [NLTK](https://www.nltk.org/) )