Fix add_class_trait in the presence of subclasses

[mdickinson]

This PR fixes the issue described in #1460, and adds some tests for add_class_trait. Prior to this PR, there were no tests at all for add_class_trait.

The PR also does a drive-by fix of a bug involving prefix traits and an invalid use of sort, leading to the error TypeError: sort() takes no positional arguments.

Unfortunately, one of the newly added tests is failing as part of the full test suite, but passing standalone. It appears that we have a test interaction - some test somewhere in the test suite is mutating the class traits on HasTraits. I'll try to track it down.

Checklist

• Tests
• Update API reference (docs/source/traits_api_reference) N/A
• Update User manual (docs/source/traits_user_manual) N/A
• Update type annotation hints in traits-stubs N/A

[mdickinson]

I'll try to track it down.

The culprit is test_abc.py, in combination with issue #58:

traits/traits/tests/test_abc.py

Line 33 in 249d648

HasTraits(x=10)

[mdickinson]

The documentation build issue that's causing the CI to fail on this PR should be fixed in #1462.

[corranwebster]

While orthogonal to main idea of the PR, there aren't tests for the is_subclass behaviour where there is an error as described in the new docstring for _add_class_trait.

Otherwise LGTM.

[mdickinson]

there aren't tests for the is_subclass behaviour where there is an error

Yes, that should be fixed. (Though the use of TraitError here bugs me: it feels like the wrong exception, and writing a test for it bakes in that wrongness a little bit harder. That's not a good excuse not to write that test, though.)

[mdickinson]

writing a test for it bakes in that wrongness a little bit harder

Though I guess I've already done that by documenting it. :-(

[mdickinson]

there aren't tests for the is_subclass behaviour where there is an error

Tests added.