Exclude display=NONE nodes from layout

[HalfWhitt]

• Fixes Display = NONE doesn't work like CSS #3322
• Fixes Confusion around style properties which accept NONE values #1502... I think. There are a number of things going on there, but if I'm reading it correctly, I think this is the last part remaining.

I've parametrized display into the existing tests for visibility, and added a new test to verify that a non-displayed widget is removed from the layout. I'm not entirely sure if it's sufficient, or if there are layout situations that could behave differently and should be tested as well.

PR Checklist:

• All new features have been tested
• All new features have been documented
• I have read the CONTRIBUTING.md file
• I will abide by the code of conduct

[HalfWhitt]

@freakboy3742 Can you think of other ways this can/should be tested, or does this seem good enough?

[freakboy3742]

The implementation looks pretty good, and I think you've got most of the bases covered as far as testing is concerned. However, I've got 2 suggestions for additional tests (or clusters of tests) that I think would be worth it.

Firstly - do display and visibility play nice together? We've already had (and test for) issues where a grandchild of a non-visible widget is set visible=True; what happens when the grandparent is display=None and I set the grandchild visible=True? I think the logic works as written, but it's the sort of edge case that warrants an explicit check.

Secondly - would it be worth having a live check of this? test_display_none proves that the math is correct; but that's all theoretical. What we really care about is whether widgets are actually in the right place, and are visible/non-visible as appropriate. #2447 is effectively a manifestation of this - a case where the implementation of visibility potentially undoes all the good work done by Pack.

If it turns out that there is an issue with the one or more backends, then we don't have to include the "live" test in this PR (or, we can include it, and skip it on the platforms where there are known bugs); but a live test would give us a lot more confidence that the style language is doing the right thing.