fix(broken-links): prefer parent-path over homepage fallbacks and block record-ID sibling suggestions

[jsanchez07]

Problem

When Mystique cannot access site content (e.g. no scraped pages cached for the domain),
it falls back to suggesting the site's homepage. That suggestion passes all existing
validation (same domain, returns 200) but gives users no useful redirect target — it just
dumps visitors at the root.

A second related problem: for broken URLs with numeric or prefixed-numeric slugs
(/event-schedule/917, /events/ticketevent-1226947352), Mystique sometimes suggests
a sibling record at the same path depth. Those records are equally deleted — or unrelated
— so the suggestion is wrong regardless of how good the path overlap looks.

Changes

src/utils/url-utils.js

• looksLikeRecordId(slug) (new private helper): returns true for pure-numeric slugs
  ("917") and prefixed-numeric slugs ("ticketevent-1226947352").
• isEntityReplacementSuggestion: extended to cover record-ID broken slugs.
  • Same-depth sibling → rejected (e.g. event-schedule/917 → event-schedule/2671).
  • One-level-up sibling → rejected unless it is the immediate parent directory itself
    (e.g. events/babymetal rejected, events/event-schedule/ kept).
  • Also relaxed the depth-equality guard so the record-ID path can be checked before
    the person-entity logic.

src/broken-links-guidance/guidance-handler.js

• nonRootSuggestedUrls filter: after filterBrokenSuggestedUrls, drops any URL
  whose pathname is / or "". Mystique's homepage fallback passes the existing HTTP
  check (200) but is never a useful redirect target.
• Parent-path fallback chain: when nonRootSuggestedUrls is empty:
  1. Prefer suggestions saved by a previous run (no change — existing logic).
  2. Try resolveParentPathFallback — walk up the broken URL one level at a time and
     return the first parent that responds 200. Never tries the root.
  3. If no live parent found, restore Mystique's original homepage suggestion
     (may be intentional — entire section removed).
  4. Last resort: effectiveBaseURL.
• aiRationale clearing: when a homepage is replaced by a parent path, Mystique's
  rationale (which was written about the homepage) is dropped so it does not mislead
  users about why the parent path was chosen.

Safety properties

• Mystique always wins — any non-homepage suggestion from Mystique is used unchanged.
• Depth-1 broken URLs are immune — resolveParentPathFallback never tries root, so
  /some-page → returns null → homepage suggestion is preserved.
• Live check required — parent path must return HTTP 200 or it is skipped.
• Existing suggestions are preserved — if a previous run stored valid URLs they are
  reused and the fallback chain is never entered.

Test plan

• resolveParentPathFallback returns first live parent, skips 404 parents, returns
  null when all parents 404 or broken URL has depth 1
• isEntityReplacementSuggestion rejects same-depth and sibling-of-parent record-ID
  suggestions; keeps the immediate parent directory suggestion
• Guidance handler uses parent-path when Mystique returns only homepage
• Guidance handler restores homepage when no live parent found
• aiRationale is cleared on homepage→parent-path replacement, preserved otherwise
• Existing suggestions from a prior run short-circuit the fallback

[github-actions]

This PR will trigger no release when merged.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[noruiz]

This also fires when we fell back to existing suggestions from a prior run which wipes existingData.aiRationale even though it's still valid. Consider moving aiRationale = undefined into the if (parentFallback) branch instead.