fix(qc,#11601): separateurs --- en *** — tranche QC/Python yaml_block_open_no_close (6 notebooks, 43 cellules)

[jsboige]

Grain: LIGHT/notebook-python — lane myia-po-2024:CoursIA — prev: LIGHT/readme #15811

See #11601

Le defaut et son remede nomme

Le body de #11601 (mise a jour du 2026-08-30) nomme la remediation : les separateurs --- en tete de cellule markdown ouvrent un bloc yaml_metadata_block non ferme au sens Pandoc — prose et titres des cellules suivantes sont lus comme paires cle/valeur, quarto render echoue, et un seul notebook fait tomber tout le site (incident #11451). Le remede outille du depot (RULE_REPAIR de detect_markdown_rendering.py, #12089/#12109) est la conversion --- -> *** (meme rendu <hr>, aucune semantique YAML).

Cette PR execute la tranche QC/Python de cette campagne, sur mesure le pool de l'umbrella.

Perimetre (6 notebooks, 63 conversions)

Notebook	Conversions
QC-Py-04-Research-Workflow	6
QC-Py-11-Technical-Indicators	12
QC-Py-12-Backtesting-Analysis	7
QC-Py-31-Transformer-Training	18
research/research_classification	11
research/research_lstm	9

Mesure avant/apres (detecteur du depot, meme arbre)

Mesure	Avant	Apres
yaml_block_open_no_close sur QuantConnect/Python	43	0
Lignes du diff	—	63 insertions / 63 deletions, toutes "---" -> "***" (audit git diff -U0 : 59x `"---
"->"***
", 4x "---"->"***"` fin de liste, zero autre ligne touchee)

Modifs uniquement markdown : exception C.2 (pas de re-execution requise), aucune cellule code touchee, execution_count/outputs inchanges (H.3 passe). Les 10 hooks pre-commit passes (gitleaks, probeAddresses, papermill paths, auto-fix yaml, oversized, source-newlines, H.3, #13326).

Le fixer garde ce qu'il ne doit pas toucher : --- dans les blocs de code, soulignements setext (titres H2), vrais blocs frontmatter.

Hors perimetre, declare

• 3 oversized_hint restants dans QC/Python (QC-Py-02 cellules 12/19/26, sections ### Indices) : regle differente, choix pedagogique a arbitrer, pas une conversion mecanique.
• Recensement repo-wide au moment de la PR : 255 yaml_block_open_no_close restants HORS QC/Python (GameTheory, GenAI Audio/Image/Video/Texte, IIT, Probas, CaseStudies...). Cette tranche ne clôt pas la campagne — familles restantes a trancher par familles.
• Claim pool-QC po-2023 du 2026-09-04 (densite round 3) : perime >48h sans livraison, perimetre distinct (enrichissement vs reparation) — note au claim #issuecomment-5649383422.

G-VAR-1, declare honnetement

Requalifie LIGHT apres recensement : une tranche de famille via fixer outillee, avec 255 instances scannables sur les familles suivantes et un hook pre-commit qui auto-fixe la regle, repond OUI au litmus « generable en scannant l'instance suivante ». notebook-python est un genre CONTENU mais le tier ne tient pas le plancher G-VAR-1 — declare, pas maquille. Budget G-VAR-2 : 0 grain merge ce jour pour la lane, budget 1, cette PR = 1. Genre precedent LIGHT/readme (#15811) : pas de run G-VAR-3.

🤖 Generated with Claude Code

[github-actions]

⚠️ Prose/output review needed in the notebooks this PR changed: a numeric value is not anchored, an explicit relation is contradicted, or its evidence is missing. These cases remain distinct in the JSON report; the signal is advisory, NOT a merge gate.

Scope = notebooks CHANGED in this PR, not the whole corpus. Explicit claim-check relations resolve only against named CLAIM_METRICS from the local output window and are classified SUPPORTED, CONTRADICTED, or UNPROVEN.
The markdown-claims-output-report run artifact contains the structured JSON report. See python scripts/check_markdown_claims_output.py --help for re-running locally.
Detector rationale: c.290 / c.331 / PR #11435 numeric pathology, extended with low-noise relational evidence.

[github-actions]

Notebook PR Validation: PASS

• Notebooks checked: 6
• Code cells validated: 131
• Result: All passed

Checks: H.1 (no errors), H.3 (execution_count), C.1 (no banned patterns)
Non-Python kernels (.NET/Lean): C.1 + errors only (execution_count advisory)
QuantConnect notebooks: C.1 + errors only (require QC Cloud for execution)

[github-actions]

Notebook outputs-required (H.4 schema): PASS (every code cell carries an outputs: list)

[github-actions]

Golden-Set Execution (H.7 P3)

✅ 8/8 notebooks passed (certified reproducible)

Notebook	Status	Time
2.1-Workflow-ML.ipynb	✅ SUCCESS	9.6s
2.2-Descente-de-gradient.ipynb	✅ SUCCESS	8.5s
2.3-Regression-lineaire-logistique.ipynb	✅ SUCCESS	12.6s
2.4-Arbres-Forets-Ensembles.ipynb	✅ SUCCESS	9.5s
Search-01-StateSpace.ipynb	✅ SUCCESS	6.8s
SL-1-LogicalLearning.ipynb	✅ SUCCESS	4.9s
rl_4_multi_armed_bandits.ipynb	✅ SUCCESS	48.8s
GameTheory-04c-NashExistence-Python.ipynb	✅ SUCCESS	6.7s

Pinned lockfile: scripts/notebook_tools/golden_set.lock.txt (H.7 P3, axe A #4208)

[jsboige]

DWELL floor justification (lane myia-po-2024:CoursIA): this PR is blocked ONLY by the PR-gate 120-min timer — head pushed 2026-09-12T23:26Z, floor elapses ~01:26Z. CI is green (74 checks, 0 failing at 23:40Z). Not repairable faster by this lane: the timer is time-based, gh pr update-branch is deliberately NOT used (resets the floor). The hourly stale-sweep (or a clean gh run rerun after the floor) will re-aggregate.