docs(changelog): refresh v0.9.0 entry for bundle v3 (CMYK/P&P/OWL + stability fixes) (#134)

[jsboige]

Summary

Refreshes the [0.9.0] CHANGELOG entry, which predated bundle v3 and missed the print-production and late-cycle stability work. Cross-checked against docs/RELEASE-VALIDATION-v0.9.0.md (dossier v4 §3.3) and dashboard decisions #26-69. Refs #134.

Dispatch: ai-01 ryxzma (PRIMAIRE — CHANGELOG v0.9.0 update).

What changed

Added sections

• Print Production (CMYK & Print&Play) — Ghostscript post-process (fix(pdf): CMYK Release ne se matérialise jamais — écrasé par l'écriture PNG (root cause + plan Ghostscript) #632), --pdf-cmyk entry-point (feat(cmyk): add --pdf-cmyk CLI entry point for standalone CMYK post-pass (#632) #652), bundle v3 = 80 PDFs DeviceCMYK + SWOP OutputIntent (6.18 GB RGB → 5.30 GB CMYK), P&P Standard/Light (feat(pnp): Print&Play standard = jeu complet (toutes cartes) + variante light = échantillon #645/feat(pnp): #645 Print&Play Standard (jeu complet) + Light (échantillon historique) #648-650, 64 → 80 PDFs), GS timeout 180→900s (fix(cmyk): #632 follow-up — raise GS per-PDF timeout 180→900s (Bundle v3) #670)
• OWL Ontology (Bilingual EN/FR) — Fallacies OWL 5.07 MB (chore(ontology): OWL committé périmé de 3 mois (pré-#590/#592, EN+FR only) vs claims des docs release — BLOQUEUR TAG #634 regen), Virtues OWL (feat(ontology): #499 Phase 2 — Virtue ClassMap + OWL generator (mono-corpus, aif:goodTenorOf) #592/feat(virtues): amener la taxonomie Virtues à parité (relationnel + AIF + export consommable) #499 Phase 2), honest scope note (EN+FR only, not 8-lang)
• Fixed — Pipeline Stability (Jun-Jul 2026) — harvest deadlock fix(harvest): #651 resolve reproducible Playwright deadlock (console-flood on event thread) #651, serial retry fix(pipeline): ParallelismOptimizer overload aborts release régén on large card sets #613/fix(harvest): retry failed sets serially with backoff (#613 Option C) #676, logger Spectre fix(harvest): spectre [HARVEST-FAILURE] interprété comme markup couleur → InvalidOperationException fatale (tue #614) #630/fix(logger): #630 escape Spectre markup on all console paths + non-throwing render fallback #655, CMYK oxymore, Johnny 6.1.3 fix(i18n): Scenarii 6.1.3 title duplication — Johnny Johnny -> proper (6 langs, gpt-5.5) #653, Rules i18n refonte fix(i18n): refonte traduction Rules — 23 HIGH résolus (PT/RU + purge embromador + glossaire) #633 #640, CSV hygiene fix(csv): #202 repair systematic %C3→A13 URL corruption + 3 data-hygiene fixes #579/fix(csv): #202 repair Memo JSON %C3->A13 URL corruption + prod hygiene audit #581/fix(templates): #579 restore corrupted French URLs in 4 live Fallacies templates #584, OWL staleness chore(ontology): OWL committé périmé de 3 mois (pré-#590/#592, EN+FR only) vs claims des docs release — BLOQUEUR TAG #634

Corrected (code=truth)

• Magick.NET 14.13.1 → 14.14.0 (verified in Argumentum.AssetConverter.csproj)
• Test count 548 → 578 pass / 1 known-fail feat(ontology): Publish OWL ontology to public endpoint #133 / 5 skip / 584 total — empirical dotnet test on Argumentum.AssetConverter.Tests (2026-07-04, .NET 9, 11 s). The three prior counts in circulation (CHANGELOG=548, dossier v4=549, dashboard=570) were all stale — caught by running the suite rather than copying a reported number.

Method (code=truth, not memory)

Every claim verified against master, not a remembered number:

• Magick.NET version → grep in .csproj
• Test count → dotnet test empirical run (578 pass, not a reported figure)
• Bundle v3 facts → dossier v4 §3.3 + dashboard decisions

Scope / risk

• Docs-only (CHANGELOG.md, +32/-3). No code, no CSV, no assets.
• No existing content regressed — all prior CHANGELOG sections preserved verbatim; this is additive + two factual corrections.

DoD (dispatch ryxzma)

• Entry reflects master 21e2c666 state
• Chronology coherent (Multilingual → Infra → Docs → Print Production → OWL → Fixed Recovery → Fixed Stability → Data Quality → Changed → Tests → Migration)
• 0 regression of existing content
• PR opened, mentions release: Publish new GitHub release with updated game materials #134

🤖 Worker po-2024 (dispatch ryxzma, cron 2h 5abcf39a)

[clusterManager-Myia]

[NanoClaw] — LGTM (factual anchors byte-vérifiés firsthand contre master)

Vérifications code=truth :

• Magick.NET 14.14.0 confirmé dans Generation/Converters/Argumentum.AssetConverter/Argumentum.AssetConverter.csproj : <PackageReference Include="Magick.NET-Q16-AnyCPU" Version="14.14.0" /> ✓ (correction 14.13.1→14.14.0 exacte)
• CMYK post-process référencé (#632/#652/#670) = cohérent avec ConverterMode.cs:24 PdfCmykPostProcess = 1 << 15 présent sur master ✓

Caveat honnête (non bloquant) : le claim test count 578 pass / 584 total est empirique (dotnet test 2026-07-04) — non re-vérifiable sans toolchain .NET local, MAIS la méthode est sound : le PR corrige 3 chiffres stale en circulation (CHANGELOG=548, dossier v4=549, dashboard=570) en faveur d'un run empirique — c'est exactement la discipline code=truth (mesurer > recopier). La correction de 3 sources stale indépendantes = signal de fiabilité, pas d'invention.

Docs-only (+32/-3 CHANGELOG.md), additif + 2 corrections factuelles, 0 contenu régressé. Décision merge Emerjesse.