Add parser ground truth tooling and arXiv dataset

[AymanL]

Summary

• Parser ground truth tooling — two new scripts under data_collection/:
  • ground_truth.py — queries arXiv API to collect verified reference articles (LaTeX source = clean ground truth, no PDF extraction artifacts)
  • download_ground_truth.py — downloads PDFs and extracts .tex source text in parallel, producing a local verified_ground_truth_data/ corpus for parser benchmarking
• verified_ground_truth.csv — committed seed dataset of 33 arXiv articles across vaccine/autism and control categories
• Tests & factories — expanded model tests and added IngestionRun / ParsedArtifact factory stubs

Test plan

• Run pytest tests/ingestion/test_models.py — all model tests pass
• Apply migrations cleanly on a fresh DB (manage.py migrate)
• Run python -m eu_fact_force.ingestion.data_collection.ground_truth --vaccine-limit 5 --other-limit 5 and confirm CSV output
• Run python -m eu_fact_force.ingestion.data_collection.download_ground_truth --csv verified_ground_truth.csv --output-dir /tmp/gt_test --workers 2 and verify PDFs + .txt files are written