Document authorisation and caching guidelines for AI features
Background
As other groups starts work more with AI powered features question start to arise around caching and authorisation (eg: !127261 (comment 1492056923))
Goal
Update documentation at https://docs.gitlab.com/ee/development/ai_features.html#features current policy and guidelines around:
- How caching works for GitLab Duo chat
- How cache AI features that are independent from the chat (eg: summarise issue comments etc)
- If authorisation should be included into cache or would it be enforced on the other layers