Clear local baseline dir in hereby test

[jakebailey]

This fixes some issues, but not the flow of using updatefailing and baseline-accept, since those affect reference baselines.

[Copilot]

Pull Request Overview

This PR adds a cleanup mechanism to the test workflow by clearing the local baseline directory before running tests, with an option to skip this cleanup when needed.

Key changes:

• Introduces a --dirty flag to control baseline cleanup behavior
• Clears local baselines by default before test runs to prevent stale results
• Extracts baseline path constants for reusability

[andrewbranch]

Thank you!