🐛(ingestion-infrastructure) rename columns for ConcoursCleaner

[AntoineAugusti]

📝 Description

Fixes #510

Rename columns when loading CONCOURS as column names have changed.

🏷️ Type of change

• 🐛 Bug fix
• 🎢 New feature (non-breaking change that adds functionality)
• 🥁 Breaking change (a modification or feature that would cause existing functionality to stop working as expected) requiring a documentation update
• 📚 Documentation update
• ♻️ Refactoring
• 🔧 Technical change

🔧 Changes

List the main changes

🛸 Required dependencies for this change (if applicable)

🏝️ How to test (if applicable)

Steps to reproduce or test

📸 Screenshots (if applicable)

✅ Checklist

• 💅 I have added or updated the appropriate tests.
• 📝 I have updated or added the necessary documentation.
• 🚀 I have considered the impact on performance, security, and user experience.
• 👀 I have requested a review from a team member.

[AntoineAugusti]

./bin/manage clean_documents --type CONCOURS
direnv: loading ~/Documents/csplab/src/tycho/.envrc
INFO 2026-05-12 09:08:53,044 [ingestion] Enqueuing clean task for CONCOURS...
INFO 2026-05-12 09:08:53,044 [ingestion] Starting cleaning 1000 document type: CONCOURS
INFO 2026-05-12 09:08:53,881 [ingestion] ✅ Clean completed: 579/4832 documents of type CONCOURS cleaned
WARNING 2026-05-12 09:08:53,881 [ingestion] ⚠️ 4253 errors occurred
INFO 2026-05-12 09:08:53,881 [ingestion] ✅ Task enqueued successfully.

---

The behavior is correct. The 4253 "errors" aren't real errors — they're rows that were intentionally filtered out by the cleaner:

Filter	Dropped
Statut != VALIDE	79
Année de référence ≤ 2024	3,985
Ministère des Armées	57

After those filters, 711 rows remain. The deduplication (grouping by concours_id) and any null category rows bring it down to the 579 entities.
The "error" count in the use case is just len(raw_documents) - len(cleaned_entities), which conflates filtered rows with real failures.

The column name update is working correctly — the DB already stored raw_data with French column names, and the cleaner now reads them properly.