Support ref in constraint and required messages with single language

[lognaturel]

Closes #474

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

Discussed approach in the issue and @MartijnR agreed identifying ${} references and forcing itext creation in that case was our best bet.

In terms of implementation, I tried to reduce redundancy in this section but I couldn't come up with any approach that would be easy to read. I think this is ok because the conditions are linear and read well.

I considered adding tests for relative references in repeats but I think that's sufficiently covered elsewhere and should Just Work.

What are the regression risks?

Everything I added is in a conditional so I think risks are minimal. The worst risk is probably that I would have forgotten some cases.

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

No, this is a bug fix.

Before submitting this PR, please make sure you have:

• included test cases for core behavior and edge cases in tests_v1
• run nosetests and verified all tests pass
• run black pyxform to format code
• verified that any code or assets from external sources are properly credited in comments

[MartijnR]

Thanks!

Looks great to me. Did some real-life testing.