masonry: Refactor local transform as Property

[Philipp-M]

This got a little bit larger than I would've hoped, and I'm not entirely sure about it (more code without immediate benefit, apart from making the transform a property, which possibly allows other APIs e.g. in Xilem).

It introduces a new marker trait GlobalProperty with impl<P: GlobalProperty, W: Widget> HasProperty<P> for W {} as suggested in #1278
And a core property Transform which is a new type wrapper around kurbo::Affine, it could reasonably also be a directly Affine (less boilerplate for sure), but I guess a new type probably is worth it non-the-less.

This unfortunately disallows other things I had in mind, due to orphan rules (this would result in a similar behavior as xilem_webs OneOf):

impl<
    A: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    B: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    C: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    D: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    E: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    F: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    G: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    H: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    I: Widget + FromDynWidget + HasProperty<P> + ?Sized,
    P: Property,
> HasProperty<P> for OneOfWidget<A, B, C, D, E, F, G, H, I>
{
}

So another case why I'm not sure about merging this. But as it's basically finished, it's at the very least worth opening this.

[Philipp-M]

Not sure whether it's the right thing to change this. Alternatively we could leave the transform_changed flag within WidgetState additionally, and check both in the compose pass (or set transform_changed as well in WidgetMut::property_changed).

[DJMcNab]

I think this alternative is the better way around. My understanding is that the semantics of prop_has_changed are very clear, which is that they mark whether the transform has changed during the lifetime of this WidgetMut, right?

[DJMcNab]

I'm finding this a bit hard to reason about.

[DJMcNab]

Suggested change

	/// When properties of a widget were modified this should also be reflected by [`ChangedProperties::mark_as_changed`]
	/// When properties of a widget were modified this should also be reflected by [`ChangedProperties::mark_as_changed`].

[DJMcNab]

Suggested change

	/// Mark that the property `P` has changed
	/// Mark that the property `P` has changed.

etc.

[DJMcNab]

Suggested change

	/// The local transform of a widget
	/// The local transform of a widget.

[DJMcNab]

I think this alternative is the better way around. My understanding is that the semantics of prop_has_changed are very clear, which is that they mark whether the transform has changed during the lifetime of this WidgetMut, right?