enhancement: finish critical-fail scoring beyond robots-blocked pages

[BraedenBDev]

Follow-up from docs/issues/2026-04-12-accuracy-from-live-audit.md (R4) and docs/plans/2026-04-18-live-audit-follow-through.md.

Problem

The audit proposed several critical-fail rules so catastrophic problems cannot hide behind a decent average score. Today, robots-blocked accessibility appears to be handled, but the rest of the proposed short-circuit rules are still not clearly implemented as a complete rubric.

Goal

Finish the critical-fail rubric so obviously broken visibility states cannot receive flattering grades.

Remaining rules to implement

• Non-JS bot content collapse:
  • if a non-rendering bot sees <200 words where Googlebot sees >1000, force an F on content visibility for that bot
• Missing required structured data:
  • if the page type expects schema and zero valid JSON-LD blocks are present, force an F on structured data
• Canonical host/domain failure:
  • if canonical is missing or points to a different domain, force an F on technical signals
• Extreme latency:
  • if TTFB exceeds the defined hard threshold, cap technical signals at D or worse

Acceptance criteria

• Each remaining critical-fail rule is explicitly encoded in scoring logic
• Fixtures exist for each critical-fail condition
• Tests assert that these cases cannot receive A/B grades through averaging
• Output explains when a category score was short-circuited by a critical-fail rule

Why this matters

Average-based scoring is fine until it flatters a broken site. Critical fails are how crawl-sim keeps its backbone.