feat: tutorial — ML-driven photo import (future/deferred)

[shaoster]

Problem / Motivation

Users with large backlogs of photo archives could benefit from an ML-assisted import path: upload photos grouped by piece name, let the system read EXIF timestamps and metadata to infer states and pre-populate fields, then review and confirm the results. This tutorial would onboard users to that flow once it exists.

Status: Out of scope — pending ML pipeline. This issue is intentionally deferred. It is included in Milestone #5 so the tutorial flow is planned end-to-end, but implementation is blocked until the ML photo-import pipeline is built.

Proposed Solution (Future)

Implement the "ML-driven photo import" tutorial using TutorialOverlay. Tutorial slug: ml_photo_import.

Steps:

1. Copy: "Have a folder of photos for an old piece? Try the Import Old Piece flow."
2. Spotlight the import entry point. Copy: "Upload photos grouped by piece name."
3. Spotlight the EXIF-derived state suggestions. Copy: "We read photo timestamps and metadata to guess your piece's timeline — review and adjust the pre-populated states."
4. Spotlight the field confirmation UI. Copy: "Confirm or correct each suggested field value before saving."
5. Completion card with Dismiss.

Acceptance Criteria (Future)

• Tutorial fires when ml_photo_import is not dismissed and the import feature flag is enabled.
• Each step spotlights the correct element.
• Dismiss marks ml_photo_import via the user settings API.
• Uses TutorialOverlay.
• Covered by unit tests.

Dependencies

• Blocked by the ML photo-import pipeline (does not yet exist).
• Blocked by overlay shell issue (TutorialOverlay component).
• Blocked by feat: add per-tutorial dismissed/seen flags to user settings #453 (dismissed/seen flags API).