Feature/make article retrieval great again

[bernomone]

• Add argparse to evals so that we can select only some models/evals to run
• Make article.py an agent again
• Fix a test with a non-existent article and made it "fuzzy" (it is enough if the title has some keywords)

Article evals passed:

Models: claude-aws-bedrock
Evals: article

===== Model: claude-aws-bedrock =====

Running article evals...
Evaluating case: Tell me about paper 'Entity Embeddings of Categorical Variables'
Evaluating case: What is paper 'The deterministic Kermack-McKendrick model bounds the general stochastic epidemic' about?
Evaluating case: Tell me about paper 'https://arxiv.org/pdf/1604.06737'
Evaluating case: What is paper https://arxiv.org/pdf/1602.01730 about?
Evaluating case: Find this paper 'Quark Gluon plasma and AI'
Total cases: 5
✅ Passed: 5

[martinapugliese]

grande Berna