fix(doc-slop): require affirmative first-time-reader checks

[shunk031]

The question is whether a researcher can understand a report on first reading with no project context. The method is to read only the report text and examine its opening, terminology, counts, section titles, and reader focus. Here, English pidgin means English concept nouns inserted into Japanese prose without a Japanese gloss. The result is that a clear report is readable, while English pidgin and audit or process narration are not. The consequence is direct: unclear reports need rewriting before publication.

What must a clear report explain?

• Its opening states the question, method, result, and consequence in plain language.
• Its Japanese terms and units are defined, and every count names its object.
• Its section titles explain what question each section answers.
• It explains the result to the reader instead of narrating audits, repository mechanics, or auditor instructions.

Which examples are clear?

• A report that leaves an English concept noun in Japanese prose is not clear and needs rewriting.
• A report with a plain opening and defined terms is clear.
• A report whose title names an internal stage and hardware mode instead of its question, and whose prose addresses auditors, is not clear.

What action follows the result?

Publish a report when a new reader can follow its question and evidence in plain language. Otherwise, rewrite the named passage until the report explains the subject to that reader.

How should this change be coordinated?

A sibling branch, feat/textlint-markdown-gate, also modifies scripts/doc_slop_review.py in the deterministic tier. This branch changes only the judge-side invocation and result parsing. The changes should reconcile when merged; whichever branch lands second must update its branch and re-verify all checks on the resulting head.

Merge-order caveat

PRs #658 and #661 also touch scripts/doc_slop_review.py and remain open. This PR changes only the judge-side invocation and result parsing; those sibling branches will update-branch after this PR lands, and the branch that lands second must re-verify its checks on the resulting head.