feat: add cv2 capture videos and images for vision llm to inference

[weedge]

feat:

• add cv2 capture videos and images
• add llm_transformers_manual_vision_qwen and test_transformers_v_qwen

  
  # gpu
  LLM_MODEL_NAME_OR_PATH="./models/Qwen/Qwen2-VL-2B-Instruct" LLM_DEVICE=cuda 
  LLM_CHAT_HISTORY_SIZE=0 \
  VIDEO_FILE="/content/capture6_2024-09-15_20-10-33.mp4" \
  python -m unittest test.core.llm.test_transformers_v_qwen.TestTransformersVQwen.test_chat_completion_prompts

LLM_TAG=llm_transformers_manual_vision_qwen LLM_DEVICE=cuda \ LLM_MODEL_NAME_OR_PATH="./models/Qwen/Qwen2-VL-2B-Instruct" \ LLM_CHAT_HISTORY_SIZE=0 \ python -m unittest test.integration.processors.test_vision_processor.TestVisionProcessor

- add pyav for video process

fix: 
- mv llamacpp_v_processor to vision_processor and add test