docs: parser-misparse-backlog.md aggregate metrics don't reconcile with its taxonomy (pre-existing)

[dan-blanchard]

Surfaced by CodeRabbit while reviewing #7047, script-verified there: docs/parser-misparse-backlog.md's aggregates disagree internally, and the gaps pre-date that PR in full (verified at base 14b11cf5b, where the same deltas appear shifted by exactly the one card #7047 removes):

• Header "Total card appearances" 4752 vs ranked-table sum 4732 vs section-heading sum 4712 vs actual bullet-count sum 4728 (at base).
• Root cause chore: update coverage stats and badges #29 is missing from both the ranked table and the sections.
• Table-vs-section count disagreements on root causes 19 (67/55), 27 (20/16), 30 (10/7), 31 (5/4).
• Heading-vs-actual-bullet miscounts on causes 1, 2, 3, 5, 22 (e.g. §2's heading said 590 while listing 597 bullets, at base).

Since the file is the canonical §3.1 worklist and its counts steer contributor ROI choices, a regeneration of tables, headings, lists, and aggregates from the clustering source (so the totals are auditable) seems warranted — flagging rather than fixing in a card PR, since a wholesale regeneration is maintainer/tooling scope.