Fix test prompt filtering

[olivercalder]

It seems the top-level prompt filters in scripted client sequence files were being ignored. In particular, these top-level filters are supposed to filter out which prompts should be ignored by the test rather than being matched against the expected sequence of prompts.

The PromptSequence field name for the filter has simply been filter, while the documentation and sequence files use "prompt-filter" for the top-level prompt filter as well as the prompt matching within the sequence. As a result, the prompt sequence creation never deserialized the top-level filters correctly, leaving them as None.

Rename the field to prompt_filter so it is deserializes from "prompt-filter" and matches the documentation and test files.

Additionally, various test cases and files had broken top-level filters due to the fact that none were being used or tested. Fix them, add top-level filters to the remaining tests which lacked them, and add a dedicated test that the top-level filters work by running two test snaps in parallel in different directories.

Lastly, to avoid any conflicts between test cases when running the tests which simply use the expect_single_prompt macro, add a filter to that macro which filters out prompts for the given interface and expects a parent directory when when run with the home interface, which it uses as a prefix filter. Since tests are run in serial, this may be unnecessary, but I believe some tests do (currently) leave behind running test snaps which can trigger prompts across test cases.

[olivercalder]

With the top-level prompt filter now in use, every test which uses a scripted client hangs. This suggests the filter isn't matching the expected prompts. I did check the source code and the matching code for prompt filters looks reasonable, with nothing obviously wrong. Notably, if a prompt filter has some fields absent, those fields are ignored and do not cause a match to fail.

Also, I think some tests leave behind active test snap instances running in the background, which I think could be making some tests hang (though some tests where this ought not to be the case, e.g. scripted_client_works_with_simple_matching, still hang even if run in isolation -- or at least if all the other tests which use scripted clients are marked as #[ignore]). For example, I think invalid_prompt_sequence_reply_errors leaves aa-prompting-test.create running until it times out, which may be longer than the test timeout.

Interestingly, when the tests hang, they don't automatically fail after the new TIMEOUT, though that's not particularly surprising, as the timeout is specific to integration.rs and does not appear in the scripted client code where the waiting must be occurring.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.96%. Comparing base (f15183c) to head (57757c3).
⚠️ Report is 107 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #290   +/-   ##
=======================================
  Coverage   80.96%   80.96%           
=======================================
  Files          21       21           
  Lines         725      725           
=======================================
  Hits          587      587           
  Misses        138      138           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[olivercalder]

I didn't close this???

[olivercalder]

Related: #297

CC @natibek