Acceptance tests: Write evals to file and use experiment names

[hanna-paasivirta]

Short Description

Acceptance tests now save judge results to a file and let you label runs with -E so you can keep results from different runs side by side.

Implementation Details

• This adds a third output file for acceptance test runs. It saves the judge verdicts to a .judges.txt file in tmp/, next to the YAML and response text.
• You can also pass --experiment=<name> or -E <name> to tag the output files so different runs don't overwrite each other.

AI Usage

Please disclose whether you've used AI in this work (it's cool, we just want to
know!):

• Yes, I have not used AI
• No, I have not used AI

You can read more details in our
Responsible AI Policy