Implement "gap" property for flexbox algorithm

[nicoburns]

Objective

• Implements flexbox "gap" property (fixes Implement "gap" property (part of the spec for flex layout) #66)
• Adds support for CSS grid properties to the test generation script (I had already added gap as part of this work, so it made sense to pull that in for this - it doesn't actually effect the generated tests because the fixtures don't actually include any grid properties).

Context

Yoga have recently implemented this. I pulled in their tests. They don't support percentage gaps, so we may need some more testing around that (or to remove that feature from this implementation).

This builds upon #246, which should be merged first.

[nicoburns]

Update: percentage gaps work fine if the parent node has a defined size, but they don't match Chrome/Firefox if the parent does not have a defined size.

[nicoburns]

Ok, I think I have percentage gaps with undefined container sizes working properly (matching Chrome/Firefox) now. I have to admit that this was a bit of a trial and error process, and I don't fully understand why it works the way it does. But I'm fairly to happy to merge it anyway as:

• I have generated tests that cover a few different cases
• The changes were relatively small, and they don't regress any of the existing tests.
• I can't imagine anyone actually wanting to use percentage gaps with undefined container sizes: the result is that the children overflow the container. I imagine that the vast majority of people will use fixed gaps AND also defined container sizes (i.e. they will avoid this code path in two separate ways).

This PR is now ready for review (although in practice is might be best to wait until to #246 is merged in order to get a better diff).

[alice-i-cecile]

#246 is merged! Can you rebase so I can review without losing my mind?

[nicoburns]

@alice-i-cecile rebased :)

[alice-i-cecile]

I'm very happy with this. It's great to see how far the code base has come. And thank goodness for the generated tests.

I've suggested some comments to help new readers get more comfortable with the "why" of gaps.

[alice-i-cecile]

Can we add an example demonstrating this?

[Weibye]

Also: ❤️ @nicoburns You've put in amazing work here over the last few weeks / days. Really appreciate it!

[nicoburns]

Can we add an example demonstrating this?

Example added.

But unfortunately in running said example, it became obvious that it wasn't computing the results quite correctly. I think this is a regression in the "fixed size gaps with undefined parent size" case due to the changes for the "percentage gaps with undefined parent size". Which wasn't caught because the existing tests because they were only testing with a defined parent size.

I've added another test to cover this case (which is still failing for now).

[alice-i-cecile]

Not merging until that regression is fixed, but the code itself is in good shape.

[nicoburns]

Regression fixed :)