frontend: global args, guidance, critic, dynamic max tokens

[OriHoch]

• Added global arguments available to all ckangpt commands:

  • --gpt4
  • --debug
  • --no-cache - disable caching of GPT resaponses
  • --clear-cache - clears the GPT responses cache before running

• use guidance instead of langchain

• misc. improvements to all frontend prompts

• added a critic step when getting the db query

• added dynamic max tokens limitation to the API, allowing to use the most tokens possible for responses

• minor changes to vectordb pineone provider

• fixes #19

• fixes #18