Background Jobs Stage 3-4

[UdeshyaDhungana]

Note

Medium Risk
Adds new background-jobs stages and updates jobs output matching rules (markers/format and trailing &), which can change pass/fail behavior for multiple stages and fixtures. CI test execution is also switched to a Docker-based runner, which may introduce environment-related test differences.

Overview
Adds Background Jobs Stage #3 (jd6) and Stage #4 (dk5) to validate jobs output for one and multiple running background jobs, including correct +/- markers and ordering.

Tightens and refactors JobsBuiltinResponseTestCase expectations (renames ExpectedOutputItems to ExpectedOutputEntries, requires the trailing & for Running entries while still allowing a regex fallback) and updates fixtures/scenarios to cover incorrect marker and output-format failures.

Updates stage registration (tester_definition.go), course metadata (course_definition.yml), Makefile stage list, and switches the GitHub Actions test job to run tests via ./docker_test/docker_test.sh instead of a native Go setup.

^{Written by Cursor Bugbot for commit 21336ff. This will update automatically on new commits. Configure here.}

[rohitpaulk]

Let's add failure test cases

[rohitpaulk]

"entry" here doesn't quite make sense - would just say "Output includes job number with PID." or something like that

[rohitpaulk]

Suggested change

	�[33m[tester::#DK5] �[0m�[92m✓ Received 1 entry for the running jobs�[0m
	�[33m[tester::#DK5] �[0m�[92m✓ Received 1 entry for the running job�[0m

(singular)

[rohitpaulk]

I think we're overusing the "Received" word here, maybe the word "found" is shorter and more descriptive of what the assertion is doing? "Found 2 entries matching running jobs", or just "✔︎ 2 entries match running jobs".

[rohitpaulk]

Same note as above - let's not use the word entry here

[rohitpaulk]

This is fine for now, but I can imagine a lot of confusion stemming from the usage of spaces here - if the user users a single space between each but has another mistake, it might look like we're complaining about spaces when we're complaining about something else.

Can't think of a neat way to solve this, but I think one way would be to use LLMs. We could expose this via the stage harness and find a mechanism to do restricted prompts (i.e. we have some guarantees over what kind of output they can emit). Then we could generate error messages like "missing space between - and Running"

[UdeshyaDhungana]

Noted - Adding this to linear.