Fix the bisect command to not cache results

[mdboom]

No description provided.

[Copilot]

Pull Request Overview

This PR ensures that benchmark results aren’t inadvertently reused by clearing the cache before parsing, and adds a sanity check to enforce that the “good” timing is strictly less than the “bad” timing during a bisect run.

• Clears the cached JSON load before computing each result to avoid stale data.
• Validates timing order and displays the bisect log on failure.

Comments suppressed due to low confidence (1)

bench_runner/scripts/bisect.py:127

• This new validation branch needs corresponding test cases to ensure the ValueError is raised and the log is shown when good_timing is not less than bad_timing.

if good_timing >= bad_timing: