feat(framework): Phase 2 PR 4 — approval workflow canonization (closes #67)

[montfort]

Summary

Fourth of 8 PRs implementing Phase 2. Canonizes the human-approval signal that issue #67 identified as a gap. Validated empirically in Sentinel (2 AIDECs applied the same shape locally as "Opción B" before this canonization).

What's added

3 optional frontmatter fields in all 11 templates that need formal approval (AIDEC, ETH, MCARD, ADR, DPIA, INC, SEC + China: PIPIA, CACFILE, TC260RA, AILABEL) across EN, ES, and zh-CN — 33 template files total:

reviewed_by: <reviewer-id>     # email | github-handle | DID
reviewed_at: YYYY-MM-DD
review_outcome: approved       # approved | revisions_requested | rejected

Ship as commented YAML so existing documents continue to validate without modification.

Semantics:

• The presence of review_outcome is the canonical "human has reviewed" signal.
• review_required: true is NOT toggled to false after approval — it remains as historical record of why review was needed.
• reviewed_at must be >= created (will be enforced by devtrail validate in PR 6).

DOCUMENTATION-POLICY.md (EN + ES + zh-CN):

• §2 Optional Fields extended with the 3 new fields.
• New §3.5 "Recording Approval" section explaining closure semantics, body section format (compatible with existing ## Approval tables in 7 templates), multi-reviewer convention for v1 (chronological body blocks; structured array deferred), and the CLI tooling on the way.

What's NOT in this PR

• No CLI code — this is templates + docs only. The devtrail approve command lands in PR 5, and devtrail validate --check-pending-reviews in PR 6.
• The 7 templates with existing ## Approval table sections (ETH, MCARD, SEC, PIPIA, CACFILE, TC260RA, AILABEL) keep their tables — both forms are documented as canonical, with the frontmatter fields as the machine-readable source of truth.
• Multi-reviewer structured review: array form — explicitly deferred until at least one adopter exercises a real multi-reviewer flow.

Test plan

• 33 templates received the same 3-field block in EN, ES, zh-CN (one Python-driven insertion script, idempotent).
• Existing documents validate unchanged — the new fields are commented out by default.
• 356/356 tests pass (no CLI code touched, but template loading exercised by integration tests).
• Manual smoke: read a few templates to verify the YAML comment block is well-formed.

Closes #67.

🤖 Generated with Claude Code