Mute audio for Spitzer playbacks that haven't received focus yet.

Mute audio for Spitzer playbacks that haven't received focus yet.

Android expects clients to not start playing audio until a focus
request has been approved by the OS; very rarely this request will
be denied (~0.5% according to UMA).  Blocking playback startup on
a hop to the UI thread would be unfortunate and slow, so instead
mute playback until approval is granted.

If approval is denied the player will be paused soon after, but
during that tiny window some amount of audio can play out, thus the
need to mute. This is done by always setting a volume multiplier of
zero when a delegate observer is registered and when it is paused.

A subsequent playback attempt will request focus and send the right
volume multiplier to the player, which will likely have already
started a silent playback in that time.

The new behavior is codified in a couple of existing tests as well
as some new ones in MediaSessionBrowserTest -- which I took some
time to cleanup as well (mixing friend class + for_testing() is
frowned upon).

BUG=529887
TEST=new tests, hammering the play button plays no audio.

[DaleCurtis, 2016-02-04 22:22:47]

Patchset #1 (id:1) has been deleted

[DaleCurtis, 2016-02-04 22:24:00]

dalecurtis@chromium.org changed reviewers:
+ mlamouri@chromium.org

[DaleCurtis, 2016-02-04 22:24:00]

Hey Mounir here's the idea I mentioned last time. I think it's much cleaner and
easier to implement as well as test.

Also, as an FYI, you have to use ForTests, ForTesting, for_tests, or for_testing
as a method name suffix for the PRESUBMIT script to catch invalid uses:

https://code.google.com/p/chromium/codesearch#chromium/src/PRESUBMIT.py&l=308

[DaleCurtis, 2016-02-10 17:43:51]

Mounir, ping?

[mlamouri (slow - plz ping), 2016-02-16 17:08:52]

This is a bit larger of a change than I was expecting. It might be because of
the test cleanup though.

Note that some stuff from https://codereview.chromium.org/1698933004 might help
here like the introduction of MediaSessionDelegate instead of using the boolean
for default returned value when no j_media_session_.

At first, I thought this was a good approach. However, I'm now worried that we
might end up muting the very beginning of some most sounds and depending on the
system charge and the sound length, it might be a significant portion of the
playback. Do you have insights about this?

[DaleCurtis, 2016-02-16 19:55:39]

Yeah ~70 of 98 lines are test cleanup :)

It won't be an issue if IPCs are working at a reasonable speed. I tested with
some gaming SFX clips I use for short sound testing: http://xorax.sea/gaming/
and approval comes way before the sound even starts. Rejection also comes before
current time even moves.