feat(core): report inconclusive status when all tests have execution errors

[christso]

Problem

When all eval tests fail due to execution errors (e.g., Not Found from a misconfigured model, network failures, auth errors), the run still reports as PASS or FAIL based on the score threshold. This is misleading — no actual evaluation happened, so the result should not be treated as a grading outcome.

Example:

1/3   ❌ violates-lightweight-core | default | ERROR: Not Found
2/3   ❌ violates-ai-first-design | default | ERROR: Not Found
3/3   ❌ follows-principles | default | ERROR: Not Found

This reports as FAIL (score below threshold), but the failure isn't due to low eval scores — it's because the provider couldn't be reached at all.

Proposed Behavior

• When all tests in a run have execution_error status, report an inconclusive or error exit code/status distinct from threshold failure
• The JUnit XML output should also reflect this (e.g., tests marked as error rather than failure)
• CLI should print a clear message: "All tests had execution errors — no evaluation was performed"
• Consider a distinct exit code (e.g., exit 2 for execution errors vs exit 1 for threshold failure) so CI workflows can differentiate

Acceptance Criteria

• Distinct exit code when all tests are execution errors
• Clear CLI messaging distinguishing execution errors from grading failures
• JUnit XML uses <error> elements (not <failure>) for execution errors