Remove VideoDecoder::PrepareForShutdownHack() and friends.

Remove VideoDecoder::PrepareForShutdownHack() and friends.

As of r156011 the pause/flush/stop shutdown dance is no more. Now that Pipeline immediately calls Stop() during teardown we can remove a pile of hacks that were used to signal that teardown was starting.

BUG=110228
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=156525

[scherkus (not reviewing), 2012-09-11 15:07:28]

NOTE: tentative CL!

I'm not sure if my GVD changes are right at all but it seems to work. Can you
test it out?

[scherkus (not reviewing), 2012-09-11 15:20:07]

Actually I forgot about the RTC-related VideoDecoders that are also relying on
these hacks. Removing that code is a bit difficult and we may have to ice this
CL until http://crbug.com/142988 is resolved.

[scherkus (not reviewing), 2012-09-11 15:47:09]

actually... PTAL

fixing Capture/RTCVD wasn't too bad

[Ami GONE FROM CHROMIUM, 2012-09-11 22:26:27]

\o/ !!!

What testing have you been able to do of the change on GVD?
(do I need to test something for you here before you land?)

LGTM assuming things still work!

[scherkus (not reviewing), 2012-09-12 10:55:15]

On 2012/09/11 22:26:27, Ami Fischman wrote:
> \o/ !!!
> 
> What testing have you been able to do of the change on GVD?
> (do I need to test something for you here before you land?)
> 
> LGTM assuming things still work!

I tried it out on mac but hit some DCHECKs() and such in that code path. It'd be
best if you try this out on Windows/Linux.

[scherkus (not reviewing), 2012-09-12 10:55:35]

+wjia can you take a look at my changes to RTC/Capture decoder?

[wjia(left Chromium), 2012-09-12 18:32:21]

RTC/Capture decoder lgtm.

[Ami GONE FROM CHROMIUM, 2012-09-12 20:32:43]

Looks fine on lumpy, which is not surprising, since the GVD::shutting_down_
member is effectively dead as of r156011; previously the only time it was
written was during Pipeline::Stop() and the only time it was read was during
GVD::Reset(), but now once Pipeline::Stop() is called GVD::Reset() is guaranteed
*not* to be called, so the variable becomes write-only.

LGTM++