Fixed form-control classes with checkboxes

[powderflask]

handle case where checkbox is rendered without label.

regression test added: test fails for ver. 2022.1 passes with this fix.

[powderflask]

Looks like tox envs may be out-of-date? I didn't touch any of that. I can try to fix the CI by removing deprecated versions and resolving dependency issues in tox envs if you like?

[smithdc1]

Yes. It's been a while since I've updated the supported versions matrix. If you have the capacity to look at it, that would be amazing.

[powderflask]

Happy to. I've updated the supported versions matrix, but django 5 introduced an innocuous change that breaks some of the tests: https://docs.djangoproject.com/en/5.0/releases/5.0/#forms

This change breaks 5 tests that (1) use SampleForm, which provides help_text for email field, and thus input acquires the new aria-describedby attribute; and (2) uses an html diff with a pre-rendered results file.

I can think of a couple complicated ways to fix this using conditionals based on Dj version (yuk), but hoped you might suggest a preferred method to resolve these 5 broken tests?

Everything else is working smoothly.

[powderflask]

Actually, I see you are already bifurcating some tests like this based on Dj version. So I'll copy that pattern and make the PR, you can review the code then to verify it is consistent with rest of tests.

[powderflask]

sigh... it gets worse.
Django assumes the help text element will have id="{{ field.auto_id }}_helptext" (see django/forms/field.html) whereas bootstrap3/help_text.html template has uses id="hint_{{ field.auto_id }}""

This means the new aria attribute rendered on the field itself refers to a non-existent id.
As per release notes, django then adds another attribute, aria-invalid="true". Django is also adding aria-invalid="true" to a few others fields too, not sure why.

Brute force fix required duplicating 7 pre-rendered test results and adding version-conditional logic to the 5 broken tests.
It's all working now, so I'll send a PR so you can review these changes. Just forewarning it's pretty hideous and a substantial increase in the number of pre-rendered test results that need to be maintained.

Let me know if you can think of an approach that will be cleaner, can do a little more work on this if needed.

[smithdc1]

@powderflask -- thanks for all your contributions here! 🙏