fix(logger): flush rolling buffer when executor errors

[not-matthias]

Summary

• When run_executor returns Err, the ? short-circuit was skipping deactivate_rolling_buffer(). The rolling buffer stayed active in its global static, so both the deferred profiler warning (e.g. "Walltime profiling is enabled, but failed to detect benchmarks") and the final error printed by main via log::error! ended up queued in DEFERRED_LOGS and never flushed — the user saw the box close and the prompt return with no indication of what went wrong.
• Capture the result, run deactivate_rolling_buffer() unconditionally, then propagate. finish() triggers flush_deferred_logs(), so queued warnings render above the final box, and by the time main calls log::error! the rolling buffer is None and the error prints normally.

Closes COD-2665

Test plan

• cargo r -- run --mode walltime -- ls shows the "Walltime profiling is enabled, but failed to detect benchmarks" warning and the "did not produce any CodSpeed result" error after the box closes
• cargo r -- run --show-full-output --mode walltime -- ls behavior unchanged
• Successful runs still tear the box down cleanly with a checkmark

[codspeed-hq]

Merging this PR will not alter performance

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

✅ 7 untouched benchmarks

---

_{Comparing cod-2665-showing-output-in-box-hides-error-messages (149a618) with main (9afaef7)}

[GuillaumeLagrange]

LGTM to me