test: add prime env eval E2E tests

[JannikSt]

Summary

• Adds E2E tests for prime env eval command
• Tests run against Prime Inference with the single_turn_math environment
• Tests include:
  • Successful evaluation with 1 example/1 rollout
  • Error handling for invalid model names
  • Error handling for missing environments

Test plan

• Tests require PRIME_API_KEY environment secret (already configured in CI)
• Run with: uv run pytest packages/prime/tests/test_env_eval.py -v

---

Note

Adds E2E tests for prime env eval covering a successful run and graceful failures for invalid model and missing environment.

• Tests:
  • E2E for prime env eval in packages/prime/tests/test_env_eval.py:
    • Installs single_turn_math test environment via uv pip (with cleanup).
    • Verifies successful eval against single_turn_math using model deepseek/deepseek-chat (1 example, 1 rollout).
    • Asserts failure with informative output for invalid model names.
    • Asserts failure when the specified environment is missing.
    • Runs in isolated temp directories and forwards PRIME_API_KEY from environment.

^{Written by Cursor Bugbot for commit 8d353d7. This will update automatically on new commits. Configure here.}