docs: replace text links with visual portals

[BrickerP]

What changed

• Record the approved creator-dogfood failure: standalone Markdown action rows did not make the dominant visuals feel clickable.
• Replace the active Human Zine contract with five independent spreads: Cover, Film, AI Usage, Process, and Open line.
• Define three whole-image native README anchors (Film, AI Usage, Open line), two plain images (Cover, Process), and exact ENTER/PLAY/OPEN A LINE affordance copy.
• Preserve the historical baseline while explicitly superseding the combined Artifact/text-link path and deleting the old Artifact asset without compatibility.

Why

The creator dogfood found a real primary-path failure: the visual posters read as static images and the low-salience Markdown rows became the actual route. Splitting Film and AI Usage is required because GitHub Profile Markdown cannot provide two reliable native hot zones or keyboard targets inside one image.

Verification gate

• Five-spread / three-anchor acceptance contract is documented.
• Keyboard focus, alt/image-failure fallback, exact destinations, pure-display SVGs, and no standalone action-link rows are explicit acceptance checks.
• The five-person recall test and all real visitor testing remain HOLD until the clickable portals are implemented and the exact remote Profile commit is read back.

Checks

• git diff --cached --check: PASS
• Documentation-only change; no local build or test run by contract.

No Profile implementation files are included in this PR. Do not merge until the separate Profile implementation and live readback satisfy the documented gate.