leverage shared LLM when it's available

[erwangranger]

a central LLM will be provided by the demo team.

when it's available, leverage that, and update the workshop to

• remove its GPU node requirement
• stop it from deploying its own LLM

or design a "fall-back" mecanism