[Android] Decouple AccessibilityInjector from WebContentsObserver

[Android] Decouple AccessibilityInjector from WebContentsObserver

Java-based WebContentsObservers can be relatively expensive, incurring
JNI overhead, including necessary object conversions, for each callback.
As AccessibilityInjector only relies on 2 observer methods, instead use
ContentViewCore's observer to signal the necessary events.

BUG=457748, 440134
Committed: https://crrev.com/f54cd4931b24a41ca835985db9c9c9db439fb5c5
Cr-Commit-Position: refs/heads/master@{#317376}

[jdduke (slow), 2015-02-20 16:58:55]

jdduke@chromium.org changed reviewers:
+ tedchoc@chromium.org

[jdduke (slow), 2015-02-20 16:58:55]

tedchoc@: PTAL. It looks like a few of our WebContentsObserver instances really
only care about a single method or two (VoiceSearchTabHelper...), so it might
just be simplest to consolidate the instances we have and then decide if the
proxy observer is really necessary.

[Ted C, 2015-02-20 17:38:58]

lgtm -- in this instance it seems ok, but I don't want to tightly couple objects
because we are trying to eliminate web contents observers in java land.  I would
greatly prefer we flush out fully the proxy object over that.

[jdduke (slow), 2015-02-20 18:55:59]

The CQ bit was checked by jdduke@chromium.org

[jdduke (slow), 2015-02-20 18:56:03]

On 2015/02/20 17:38:58, Ted C wrote:
> lgtm -- in this instance it seems ok, but I don't want to tightly couple
objects
> because we are trying to eliminate web contents observers in java land.  I
would
> greatly prefer we flush out fully the proxy object over that.

Sounds good.

[commit-bot: I haz the power, 2015-02-20 18:57:25]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/941353002/1

[commit-bot: I haz the power, 2015-02-20 19:45:03]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-02-20 19:45:40]

Patchset 1 (id:??) landed as
https://crrev.com/f54cd4931b24a41ca835985db9c9c9db439fb5c5
Cr-Commit-Position: refs/heads/master@{#317376}