NOT FOR COMMIT: Move background pausing to the WebView level

NOT FOR COMMIT: Move background pausing to the WebView level

This way the pausing happens per tab instead of per process.
This also means we pause the suspendable task queue as well.

BUG=709082, 670102

[Z_DONOTUSE, 2017-05-04 22:48:22]

Description was changed from

==========
move test


Deleted all the code. Just needs tests?


Delete all the dead code.


It all seems to work as intended!
==========

to

==========
NOT FOR COMMIT: Move background pausing to the WebView level

This way the pausing happens per tab instead of per process.
This also means we pause the suspendable task queue as well.

BUG=709082,670102
==========

[Z_DONOTUSE, 2017-05-05 00:20:52]

ojan@google.com changed reviewers:
+ ojan@google.com

[Z_DONOTUSE, 2017-05-05 00:20:53]

altimin/skyostil: FYI, I have a patch in progress for this. It needs tests and a
real change description, but I think is otherwise ready.

Would welcome any early feedback you have, or if you wanted to take this over
given that I have limited time to write code. Otherwise, I'll try to finish this
soon.

I actually started by trying to plumb Site Engagement for the desktop background
pausing and realized that this step should probably happen first.

[Sami, 2017-05-05 14:36:36]

skyostil@chromium.org changed reviewers:
+ skyostil@chromium.org

[Sami, 2017-05-05 14:36:36]

This seems workable, although I wonder if still need to suspend the top level
timer/loading task queues when all WebViews in the process are suspended given
that not everything is tied to a per-frame queue. I guess the UMA data from this
vs. the existing scheme will tell us if we are improving things overall.

[Z_DONOTUSE, 2017-05-05 17:27:03]

On 2017/05/05 at 14:36:36, skyostil wrote:
> This seems workable, although I wonder if still need to suspend the top level
timer/loading task queues when all WebViews in the process are suspended given
that not everything is tied to a per-frame queue. I guess the UMA data from this
vs. the existing scheme will tell us if we are improving things overall.

Hmmm...that's a good point. I don't think we need to wait for UMA. We should
suspend them as a matter of principle if nothing else. I'll try to modify this
patch to identify when all WebViews are suspended and suspend the top-level
queues as well.

Again, I have very limited coding time, so expect slow progress and don't
hesitate to take this over if either of your find yourselves with the time. I'm
not at all attached to being the one that finishes this patch. :)

[Z_DONOTUSE, 2017-05-05 20:53:28]

Ugh. Elliott also pointed out that this isn't safe as is. A foreground popup
window could be talking to a background tab. So we can only suspend tabs if the
whole siteinstance is suspsendable. I don't know if the renderer even knows
about the other WebViews in it's SiteInstance.

[Z_DONOTUSE, 2017-05-05 23:38:57]

Looks like there's an idea to plumb this data down into platform in
crbug.com/718489. I think this patch is probably blocked on that.