Guard monitoring and webhook prose, Turkish YAML comments, and Annex IV label parity #949

Review of #946 found documentation gaps invisible to currently passing guards. check_bilingual_code_blocks.py compares YAML keys, not comment language; a Turkish monitoring example containing six English comments passed. check_bilingual_parity.py compares heading-level signatures, allowing reordered same-level configuration sections to pass. Source Annex IV labels and the repeated documentation tables are not cross-checked. Webhook delivery claims in troubleshooting and wizard translations are outside the site-claims guard.

Acceptance criteria:

• Add deterministic regression coverage for the observed Turkish-comment regression and reordered configuration sections, with an explicit, maintainable source of expected correspondence rather than broad language guessing.
• Check repeated Annex IV field labels against one authoritative mapping after the legal mapping is corrected in Art. 9 paragraph citations for risk_assessment fields are wrong in config descriptions, wizard prompts, the user-manual config reference and the six-language site wizard (CR-A01-062) #539.
• Cover the monitoring/webhook claims in manual troubleshooting and all wizard locales; removing the correction must fail the relevant check.
• Document which design pages describe current behavior versus historical proposals; in particular classify the live-looking Art. 12+17 text in docs/design/wizard_mode.md.
• Wire any new guards into CI and all documented gauntlets, with negative/mutation tests.

Node wizard execution and affected-locale key parity are covered by #946; this issue covers prose/comment meaning and source-to-doc parity, not a duplicate JS harness request. Related: #670, #684, #902.