[FIX] eliminate hash collisions in test caching

[jtwaleson]

Description

In a recent PR Felienne noticed something weird: a test that should fail actually succeeded. aa3f4d9

After debugging we figured out that:

• we expect some tests to succeed
• in some tests we expect to raise an exception

When the same snippet and level is used across tests, the test runner would use a cached result from another run, reporting the wrong outcome.

Example:

• Test 1: Snippet A - Level 3 - expect successful execution -> this runs fine
• Test 2: Snippet A - Level 3 - expect an error -> this should fail because we don't get an exception BUT it does not raise an error because it re-uses the cached result from Test 1.

In this PR we include not just the snippet and level but ALL of the arguments to the single_level_tester method. This way, if we expect an exception, it will result in a different hash.

How to test

• See if all the tests still run fine -> https://github.com/hedyorg/hedy/actions/runs/6988954244/job/19017110194
• See if caching still improves the test run by a factor of about 3x -> https://github.com/hedyorg/hedy/actions/runs/6989091374/job/19017406242

Checklist
Done? Check if you have it all in place using this list: (mark with x if done)

• Contains one of the PR categories in the name
• Describes changes in the format above
• Links to an existing issue or discussion -> no, this PR describes the issue that was discussed on Discord
• Has a "How to test" section

[ghost]

👇 Click on the image for a new way to code review

Legend

[Felienne]

Thanks for the extensive detective work!!

[mergify]

Thank you for contributing! Your pull request is now going on the merge train (choo choo! Do not click update from main anymore, and be sure to allow changes to be pushed to your fork).

[mergify]

Thank you for contributing! Your pull request is now going on the merge train (choo choo! Do not click update from main anymore, and be sure to allow changes to be pushed to your fork).