Also, change the hit testing geometry so that we get our hits at the root.

We need to compute the paint bounds of ChildViewLayers during Preroll.
Also, we need to allocate a BlendPtr before writing into it.

We push a bit up the tree during preroll to learn whether there system
composited layers below each layer. During update scene, we squash down
to paint tasks, which we execute after publishing the new scene.

Example: $ mojo:launcher mojo:spinning_square

This patch removes almost all //base dependency of //flow. The only dependency
left is on tracing.

* Added instrumentation for timings.

* A few more traces for better granularity.

Instead, make Layer::Paint take a new PaintContext that has just exactly
the state that it needs, mirroring PrerollContext. Also, rename
PaintContext to CompositorContext because it holds the context for the
whole compositor.

We need to adjust the transform matrix to undo the device scale.

We weren't correctly dividing by the device pixel ratio when asking Mozart to
present our child scene.

When linking in a scene, it's a good idea to specify its clip rectangle
to ensure that its content does not draw or capture input from its
surrounding space.

A future version of the Mozart view manager may automatically wrap a
view's scene with a clip rectangle to avoid composition gotchas like this
but it currently doesn't.  So the net effect is that if you embed a view
without specifying a clip, it can draw or capture input from anywhere
within its container.

Previously we weren't factoring in the layer's offset properly and we were
drawing the child views below the background layer.

Previously we just dropped child scenes on the floor. Now we upload them
to Mozart. However, we just draw them on top of all the other content
and don't apply any clips or blends.

This patch contains Dart bindings for adding a child scene when running in
Mozart. The child scene is currently ignored, but a later patch will actually
incoporate it into the tree.