update main branch

[andreashappe]

This changes "a lot":

• add an additional query to the loop: the LLM should create a list of facts/ideas about the tested system. This fact list is then subsequently used to generate the next command to execute
• allow selection of other OpenAI models, e.g., gpt-4
• make context size configurable
• clean-up templates
• switch output to using rich for nicer looking tables/panels
• switch from python-openai to HTTP-based requests to lighten dependencies
• refactor code..