feat: show test request and response payloads

[fruffin-le]

Summary

• capture provider request and response payloads alongside test-run outputs
• add Cerebras structured-output schema support and normalise unsupported keywords
• make result payloads readable and copyable, with a failing-tests filter in the details modal
• added ability to only show failing tests in the test results modal

Testing

Before:

After:

[glesage]

LGTM other than minor comment but do you mind adding some screenshots of what it looks like now with (and without) the new feature 🙏