local-ai: Preselect Local AI for eligible systems during onboarding

Problem

Local AI is easy to overlook during onboarding because it is not selected by default and does not stand out in the flow.

@jacobtomlinson, tagging you for Local AI onboarding and product context.

Proposed behavior

Preselect Local AI when the existing qualification result marks the machine eligible, while preserving an explicit opt-out before setup downloads a model. Preselection alone must not start a download.

• Preselect Local AI only on eligible systems.
• Clearly explain the expected download size, storage use, and setup time before the user continues.
• Keep the opt-out prominent and ensure opting out skips the model download and Local AI service setup.
• Preserve an explicit opt-out across Back navigation, resumed setup, and app restart.
• Keep the current disabled state and eligibility explanation on unsupported systems.

Acceptance criteria

• Local AI is preselected only when the existing qualification result marks the system eligible.
• Preselection does not start a download without the user's confirmation.
• The user can opt out before any model download begins.
• Download and storage impact are visible before confirmation.
• Existing installations and previously saved choices are not unexpectedly changed.
• Onboarding tests cover eligible, ineligible, explicit opt-out, Back navigation, resumed setup, and app restart.

Start at the onboarding flow and trace the existing qualification result, Local AI selection, and setup download path. Review the onboarding tests and add coverage for eligible and ineligible systems, confirmation and opt-out behavior, Back navigation, resumed setup, and app restart. Done means selection and persistence follow the acceptance criteria without changing existing installations or saved choices.