Several Enhancement Items to Help Improve Repo Readability

[mingxin-zheng]

• Restructure the repository to eliminate the duplicated function calls
  For example, similar functions are duplicated in util.py and util.py
  The root cause should be that we lack modulization of the expert function

• Provide an instruction to set up local environment to run demo without Docker
  Currently, it lacks clear steps to try the models easily without Docker
  related issue: Something Error #72

• Abstract the expert model as tool calls, which may be furthered wrapped as MCP servers for other models to consume

• Update the VILA dependencies from VILA 1.5 to latest vila code to leverage new features such as remote code execution and json output

• Update the model checkpoints on HuggingFace to support remote code execution

• We may also improve the data_prepare scripts, as it seems there are several works to augment medical annotations using LLMs. We could possibly find the common aspects and provide some improvements to the current workflow.