chore: remove smoke test

[Cedric / ViaDézo1er (viadezo1er)]

Remove smoke test doing an almost empty call to claude/codex save a few seconds in

◇  How do you want to add Braintrust to your application?
│  Use built-in coding agent
│
◆  Which coding agent should Braintrust Setup use?

where the user can only wait.

I'm wondering, the code is a bit too polished but at the same time a bit too sloppy.
Is there a reason to add a smoke test or was it slop Luca Forstner (@lforst)?

If we keep them, I'll set codex effort to low instead of xhigh.

[Luca Forstner (lforst)]

Yeah it was intentional to check whether the agent will actually do something when invoked. It is also intentional that we do not use different settings than with the normal run, because we want to see in a relatively quick and reliable way whether the agent would work.

I updated the pr to use cheaper models and lower reasoning so that the check is faster. I'll leave it up to you tho whether you want to merge with this newfound information I just gave you.

[Cedric / ViaDézo1er (viadezo1er)]

Let's not change anything then, this steps takes ~7s that's fine.