docs(media-buy): correct brand requirement guidance

[bokelley]

Summary

• document that get_products requires buying_mode while brand is conditional on catalog
• preserve create_media_buy as always requiring brand
• separate buyer input-required from seller-internal async review
• replace invented policy response fields/statuses with canonical outcomes
• clarify portfolio brand_id and industries resolution
• add schema/prose regression coverage

Verification

• doc compliance drift tests: 14/14
• docs navigation: 21/21
• owned links: 20/20
• targeted snippet checks for both pages
• direct schema validation for all new response examples
• git diff --check
• independent protocol review: no remaining findings

No changeset: documentation-only guidance correction, no protocol surface change.

Closes #6457

[aao-secretariat]

Ladon verdict: Approve

Approve — docs-only correction with a schema-anchored regression test.

This PR touches two media-buy guides (docs/media-buy/media-buys/policy-compliance.mdx, docs/media-buy/product-discovery/brief-expectations.mdx) plus a new lint test (tests/lint-doc-compliance-drift.test.cjs). The reviewer verified every wire-shape claim against the schemas:

• brand optional for get_products unless catalog is present (get-products-request.json dependencies.catalog=['brand']); brand required for create_media_buy; brand.industries/brand_id override (brand-ref.json) all confirmed.
• rejected/input-required/submitted/failed outcomes with POLICY_VIOLATION + recovery:correctable + two-layer errors[]/adcp_error confirmed.
• Removes an invented policy_compliance status field that never existed on the wire; swaps real 'Nike Air Max' for fictional 'Nova Run' (satisfies the fictional-example convention); adds a schema-anchored lint.

Changeset scope: the edited paths are under docs/media-buy/, not docs/reference/** or static/schemas/source/**, so no changeset is owed. No schema files modified, so schema↔docs coherence, oneOf, and released-artifact gates do not apply. Base is main, not 3.1.x. No blocking or medium findings.

Decision path: no critical/high findings (rows 1 skipped); gated_paths false (row 2 skipped); high_risk false (rows 3/5 skipped); no medium findings (rows 4/6/8 skipped); no team gate (row 7 skipped). Falls through to row 9 → approve.