fix(cli): one guard for the two ways a run can say nothing about a model

[evkir]

Codecov found a line the patch never executed, and it was a duplicate: an empty result list falls through the emptiness check and the no-proof check to the same answer, so the first guard could not be reached except through the second.

Removed, and the reachable path it stood for is now tested instead. An external suite whose checkout is absent scores 0/0 and still writes a card -- nothing ran, so the rollup has to answer that rather than divide by an empty run. Dropping the surviving guard raises a KeyError on that path, which is what the test catches.

What this changes

How it was measured

Checklist

• ruff format --check cyberai/ tests/ and ruff check cyberai/ tests/ pass
• pytest -W ignore::DeprecationWarning -m "not slow and not smoke" passes
• New behaviour is covered by a test that fails without the change
• I have read CLA.md and I hereby sign the CLA

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!