Use origin as MediaCapabilities feature.

This CL somewhat hackily plumbs the (hashed) origin from the render
frame to the LearningHelper.

What should happen, instead, is this:

 - LearningTaskController is created per frame.
 - LearningTaskController has a FeatureProvider that memorizes the
   frame's origin.

However, since LearningHelper is created per-context, there's no
easy way to get the frame into its feature provider.  We could
plumb a separate feature provider per frame, I suppose, and let
the VideoDecodeStatsRecoder handle the lifetime, but that requires
more changes than this.

If it turns out that origin is actually useful, then we can use
per-frame LearningTaskControllers more properly once they're done.

Change-Id: I440dc5090c690865716fff5ca9a521a12e474083
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1521448
Reviewed-by: Christian Dullweber <dullweber@chromium.org>
Reviewed-by: Chrome Cunningham <chcunningham@chromium.org>
Reviewed-by: Bo <boliu@chromium.org>
Commit-Queue: Frank Liberato <liberato@chromium.org>
Cr-Commit-Position: refs/heads/master@{#642551}
14 files changed