FIX [SCMS] Make seedMarkerNote boot-idempotent + pointer-repair safe 🧬

[AdaInTheLab]

Summary

This PR fixes a long-standing mismatch between how the API marker note was seeded and how the Lab Notes system actually identifies and serves notes.

The root cause was an identity mismatch:

• lab_note_revisions.note_id is a foreign key to lab_notes.id
• The marker note already existed in dev with id = api-marker:en
• Tests and seed logic incorrectly assumed id = slug:locale (api-marker-note:en)
• This caused FK failures, missing revision pointers, and persistent 404s in the public detail route

What changed

• Seed logic now resolves the real note primary key
  • Looks up the marker note by (slug, locale)
  • Reuses the existing lab_notes.id if present
  • Only creates a new note if it does not already exist
• Revisions are now always inserted using the resolved note PK
  • lab_note_revisions.note_id correctly references lab_notes.id
• Revision pointers are repaired deterministically
  • current_revision_id and published_revision_id are always set to the latest revision
• Marker note is guaranteed to be publicly readable
  • status = published
  • published_at ensured
  • content_html populated if missing
• Lab Note detail tests now pass reliably
  • No 404s caused by identity, FK, or projection filters

Why this matters

This aligns the marker note with the system’s real identity model instead of an assumed one.
It makes seeding:

• safe in :memory: test DBs
• safe in existing dev/prod DBs
• idempotent on every boot

It also prevents future regressions where a note “exists” but is invisible due to mismatched IDs or missing revision pointers.

Notes

• The marker note intentionally uses a legacy ID format (api-marker:en)
• Seed logic must never assume lab_notes.id === slug:locale
• This PR codifies that rule instead of fighting it

Tests

• Lab Note detail › GET /lab-notes/:slug returns a note with contentHtml now passes
• No schema or migration changes required