[fuchsia][a11y] Set explicit hit regions per compositor layer on GFX.

[abrush21]

This change modifies the GFX external view embedder class to set hit regions per compositor layer, which correspond to the bounds of the paint operations performed on the underlying skia canvas. The purpose of this change is to enable more accurate accessibility hit testing of flutter overlays. Today, scenic (fuchsia's graphics compositor) assumes all overlay content is full-screen, even if the overlay only paints content in a small region. With this change, scenic has enough information to determine accurately where overlay content is located.

fxbug.dev/100825

Pre-launch Checklist

• I read the Contributor Guide and followed the process outlined there for submitting PRs.
• I read the Tree Hygiene wiki page, which explains my responsibilities.
• [X ] I read and followed the Flutter Style Guide and the C++, Objective-C, Java style guides.
• I listed at least one issue that this PR fixes in the description above.
• I added new tests to check the change I am making or feature I am adding, or Hixie said the PR is test-exempt. See testing the engine for instructions on
  writing and running engine tests.
• I updated/added relevant documentation (doc comments with ///).
• I signed the CLA.
• All existing and new tests are passing.

If you need help, consider asking for advice on the #hackers-new channel on Discord.

[chinmaygarde]

cc @akbiggs

[arbreng]

Lets go over this via GVC, there is some complex stuff I need to explain to you in here

[arbreng]

The complex part I was talking about before....

Lines 302-306 need to happen after the Present() below

We pipeline the Skia rendering work with the scenic IPC, and the fences passed to Present() make sure Scenic delays internally until Skia is finished

Moving the Skia rendering work up here will likely cause a performance regression

You really only need to move "finishRecordingAsPicture" up here, in order to populate the RTree. Can you break this part into 2 pieces, before and after Present()? It would also help to add a comment below, on the Skia code after Present, explaining the pipelining (my bad for not doing that originally)

[abrush21]

Thanks for the explanation. I broke up the pre- and post-Present() operations as you suggested.

I had to cache the sk_sp returned by finishRecordingAsPicture at the beginning of the method, so that I could draw it to the canvas after the Present() call at the end of the method. Is that safe to do?

[arbreng]

Yeah that should be safe to do.

finishRecordingAsPicture doesn't actually do any GPU work, it just records the commands from the SkRecordingCanvas into an SkPicture (a type of command buffer).

The canvas->drawPicture after Present() is what actually does the GPU work

[akbiggs]

Something I find useful for tests is to leave a comment at the top of the test explaining what behavior the test is validating since sometimes it can be hard to tell from the name alone.