enhancement: validate crawl-sim JSON outputs against JSON Schema in CI

[BraedenBDev]

Follow-up from docs/issues/2026-04-12-accuracy-from-live-audit.md (M4 remainder) and docs/plans/2026-04-18-live-audit-follow-through.md.

Problem

docs/output-schemas.md exists, but the contracts are still only human-readable. There are no *.schema.json files and CI does not validate actual script outputs against machine-readable schemas. That means output drift can still land silently.

Goal

Make output contracts enforceable.

Scope

• Add JSON Schema files for the core script outputs, at minimum:
  • fetch-as-bot.sh
  • extract-meta.sh
  • extract-jsonld.sh
  • extract-links.sh
  • check-robots.sh
  • check-llmstxt.sh
  • check-sitemap.sh
  • diff-render.sh
  • compute-score.sh
  • build-report.sh
• Add CI validation that runs representative fixtures through those schemas.
• Document the schema directory and validation workflow in docs/output-schemas.md.

Acceptance criteria

• schemas/*.schema.json exists for the script outputs above
• CI fails if fixture output no longer matches its schema
• docs/output-schemas.md points to the machine-readable schemas, not just prose examples
• At least one regression test proves schema validation catches a malformed fixture

Why this matters

The live audit showed that crawl-sim's raw data can be correct while the interpretation layer drifts. Locking the JSON contracts reduces that drift surface area.