Add support for token-based configuration to control and manage AI usage. This feature should allow defining token limits and related settings per user, project, or environment to ensure predictable costs and prevent overuse.

Acceptance Criteria:

• Ability to configure max tokens per request and per period
• Configuration can be set at project or user level
• Clear error/feedback when token limits are exceeded
• Defaults documented and configurable via config file or env vars

This will help with cost control, reliability, and safer AI adoption across the platform.