v1.10.0 — neuroslop catalog + AD-6 strengthening

Adds a model-agnostic catalogue of AI-generated-Russian-prose tells and sharpens manufactured-antithesis detection.

Added

• AD-8 — Assistant-register meta-commentary: sycophancy («Отличный вопрос!») and assistant sign-offs («Надеюсь, это помогло»).
• AD-9 — Hollow opener: «давайте разберёмся», «погрузимся», «важно понимать, что», filler «итак» — density is the primary signal.
• Neuroslop index — a named digest mapping each tell to its canonical home (no per-model breakdown).
• Table of contents for the addenda; /ru-check step 5 now scans AD-6..AD-9; README neuroslop use-case in both languages.

Changed

• AD-7 extended to positive-polarity self-praise («чётко, по делу», «коротко и ясно») + the unproven-claim tricolon.
• AD-6 strengthened: new triggers «не просто X, а Y», «вопрос/суть не в X, а в Y»; a cluster lands Structure in the 5–6 band or lower while a single antecedent-backed pair does not — stronger, but never to zero.
• version 1.9.0 → 1.10.0 across all manifests.

Every new rule carries explicit false-positive carve-outs; an adversarial false-positive review and a behavioral ru-score test passed. No correctness coverage removed.

Also closes the GitHub Community Standards gaps (Code of Conduct, Security policy, PR template, #17).