Don't do unnecessary condition evaluations in assertEventuallyTrueInGui

[mdickinson]

GuiTestAssistant.assertEventuallyTrueInGui does an unnecessary extra condition evaluation after the event loop has exited. In the worst case, when the condition is only valid during a running event loop, this can cause spurious test failures.

This PR updates the code so that it doesn't re-evaluate the condition once the condition has evaluated to True for the first time.

Design note: the bool() call is in there to avoid hanging onto references in case the user-supplied condition returns something more complicated that a simple bool.

---

EDIT: I've also updated the PR to use the exit method instead of the quit method within event_loop_until_condition; all we want here is to exit the event loop that was started using self.qt_app.exec_ - not to tear down anything else.

[mdickinson]

CI is failing for unrelated reasons. I suggest that we pin our PySide6 version in CI in the short term.

[mdickinson]

1169 should fix the unrelated CI failures, so that we can see whether there are any related failures.

[mdickinson]

Looks like this makes some existing macOS tests unhappy. Investigating.

[mdickinson]

Marking as draft, since there seem to be some unexpected interactions with the ModalDialogTester: see #1170.

[corranwebster]

The failures for the font and color dialogs seem to be happening in the tests for the utility functions get_color and get_font which wrap the dialogs, rather than the main tests for the dialogs themselves.

Possibly there is an issue with needed objects being gc'd or not cleaned up correctly?

[corranwebster]

This should resolve #1170 as well.

[mdickinson]

Thanks, @corranwebster. I'm happy for this to be merged if you are.