Allow organizations to restrict reasoning effort levels

Currently, Github copilot allows organizations to control which AI models are available to users, but it does not provide similar controls for model reasoning or effort levels.

This can lead to unexpectedly high costs because users may select premium models with the highest effort level even when that level of reasoning is unnecessary. For example, using Fable 5.1 with Max effort may be significantly more expensive than using a lower effort level for routine development tasks.

Organizations should be able to configure which effort levels are available for each model. For example, administrators could:

- Disable Max effort for selected premium models
- Allow only Low or Medium effort by default
- Define different limits for specific teams or users
- Set organization-wide default effort levels

This feature would help organizations manage AI usage costs more effectively, prevent accidental use of expensive configurations, and ensure that premium reasoning levels are reserved for tasks that genuinely require them. It would also give administrators more granular control without completely blocking access to useful premium models.