benchmark: commit0 instances with num_passed 0 and num_tests 0

[juanmichelini]

Observed on: #82 (update to commit hash once merged)

Steps to reproduce:

run
uv run commit0-infer .llm_config/sonnet-4.json --repo-split lite --n-limit 16 --num-workers 16 --max-iterations 100 --note "test-commit0-$(date +%Y%m%d%H%M%S)"

see the output.jsonl file, some intstances have "num_tests":0 and "num_passed":0

{"instance_id":"parsel","test_result":{"eval_result":{"name":"parsel","sum":0,"passed":0,"num_passed":0,"num_tests":0}}, ...}
This should be addressed on commit0/run_infer.py

Currently the instances that work are due to this fix

# Install pytest and required plugins for test reporting
        plugin_install_cmd = f"cd /workspace/{workspace_dir_name} && (uv pip install pytest pytest-json-report pytest-cov || pip install pytest pytest-json-report pytest-cov)"
        res = workspace.execute_command(plugin_install_cmd, timeout=600)
        if res.exit_code != 0:
            raise RuntimeError(f"Failed to install pytest and plugins: {res.stderr}")
        logger.info("Installed pytest and required plugins")

        # Verify pytest and plugin installation
        verify_pytest_cmd = f"cd /workspace/{workspace_dir_name} && pytest --version"
        verify_pytest_res = workspace.execute_command(verify_pytest_cmd, timeout=60)
        logger.info(f"Pytest verification exit code: {verify_pytest_res.exit_code}")
        if verify_pytest_res.exit_code == 0:
            logger.info(f"Pytest available: {verify_pytest_res.stdout.strip()}")
        else:
            logger.warning(f"Pytest verification failed: {verify_pytest_res.stderr}")

        verify_plugin_cmd = f"cd /workspace/{workspace_dir_name} && python -c 'import pytest_jsonreport; print(\"Plugin available\")'"
        verify_plugin_res = workspace.execute_command(verify_plugin_cmd, timeout=60)
        logger.info(f"Plugin verification exit code: {verify_plugin_res.exit_code}")
        if verify_plugin_res.exit_code == 0:
            logger.info("pytest-json-report plugin verified successfully")
        else:
            logger.warning(f"Plugin verification failed: {verify_plugin_res.stderr}")

Maybe they can be refactored in a function prepare_instance_test_framework and completed to work for all instances.