fix: no labels visible for randomized selects with translations

[lindsay-stevens]

Per report in https://forum.getodk.org/t/54492

• regression seems to be from Case insensitivity of sheet, column, and setting names; plus more performance improvements #746 where the choices may be referenced by the survey only and the select question does not have a reference. This happens if the question is randomized, or external select, or search appearance.
  • this would suggest that maybe the choices should be in-line for randomized choices - however the existing tests for randomize expect an itemset in the body. Also, these new tests pass on tag 3.0.1 (prior to Case insensitivity of sheet, column, and setting names; plus more performance improvements #746), and fail on master (e0b350b)
  • since the user noticed this via https://getodk.org/xlsform/ and it seems that the converter uses https://staging.getodk.cloud I assume that means master is deployed to staging (i.e. not just tagged releases)
  • The form posted in the forum converts as expected (choice labels shown in enketo preview) without warnings/errors when uploaded to a project on https://staging.getodk.cloud

Why is this the best possible solution? Were any other approaches considered?

As mentioned above I wasn't totally sure that these combinations are supported, considering enketo/enketo#322 but it seems like it works in Collect (per https://forum.getodk.org/t/42506/4) and presumably will in Webforms too. Also I wasn't able to locally reproduce the ODK Validate warnings mentioned in the reporting forum thread but maybe the Validate version on xlsform-online is out of sync?

What are the regression risks?

This should be a fix since the existing tests pass and the new code satisfies tests that worked on earlier code (as described above), and the code committed in between didn't intend to change randomize or choice_filter behaviour for users.

Does this change require updates to documentation? If so, please file an issue here and include the link below.

No

Before submitting this PR, please make sure you have:

• included test cases for core behavior and edge cases in tests
• run python -m unittest and verified all tests pass
• run ruff format pyxform tests and ruff check pyxform tests to lint code
• verified that any code or assets from external sources are properly credited in comments

[lognaturel]

Thanks for spotting this on the forum and for the speedy fix!

[lognaturel]

maybe the choices should be in-line for randomized choices

Do you mean directly as children of the question? That wouldn't work with the way randomization currently works -- it relies on an itemset declaration that references a secondary instance being able to include a function call. It's possible to randomize nodes from external secondary instances and even repeats! I think Enketo supports those, what doesn't work is if there's a predicate in the expression (e.g. [age > 17]).

master is deployed to staging

Yes, exactly. It's to catch issues like this one as early as possible. Should we add this to the readme, maybe? The only place I can think of where it's documented is here (ODK internal)

when uploaded to a project on https://staging.getodk.cloud/

Ahhh, we have various staging servers. 🙈 That one reflects the latest Central but doesn't also update pyxform. The user was using the pyxform staging server.