Improve flaky test debugging by preserving attempt-level traces (first failure + final)

Currently when a test fails and passes on retry, we mark it flaky but only show the final attempt. That hides the original failure and makes debugging harder.

Thinking of storing attempt-level traces instead, but keeping a single merged view — and just showing the first failure trace + final result.

More details here

Maybe later we can optionally expand to see all attempts if needed. Thoughts?