Fix X11DesktopHandler::ActivateWindow on Unity for known user gestures

Fix X11DesktopHandler::ActivateWindow on Unity for known user gestures

Unity in Ubuntu 14.04 ignores X11DesktopHandler::ActivateWindow's
requests to raise Chrome windows by changing _NET_ACTIVE_WINDOW. Instead
it treats it as a _NET_WM_STATE_DEMANDS_ATTENTION and just flashes the
Launcher icon.

This isn't good enough in cases where the activate request was triggered
by a user gesture, for example the user clicked on a push notification.

This patch changes X11DesktopHandler::ActivateWindow to indicate to the
window manager when it is being called from a user gesture (see the
comment added there for more details). And it plumbs the presence of the
user gesture all the way from the content/ layer.

In particular this patch fixes https://crbug.com/470830 on Linux. In
theory the plumbing will be reused when fixing the same issue on Mac.

I tried to make the fix somewhat general, by making ~ScopedBrowserShower
use the user gesture path for any existing navigation with user_gesture
set true in NavigateParams (see chrome/browser/ui/browser_navigator.cc).
Hence the patch also fixed https://crbug.com/489256 (as
chrome::ShowAboutChrome already has user_gesture set to true).

However it wasn't possible to migrate all callers of activate that are
in response to a user gesture (as there are hundreds of callers) - that
will need to be done as follow-up patches, on a case-by-case basis. For
example, this will likely allow https://crbug.com/411702 to be solved
in future.

BUG=470830, 489256, 411702
TEST=manual due to OS dependency, follow steps in https://crbug.com/470830 and https://crbug.com/489256

[johnme, 2015-05-18 13:28:48]

johnme@chromium.org changed reviewers:
+ avi@chromium.org, erg@chromium.org, msw@chromium.org

[johnme, 2015-05-18 13:28:49]

avi@chromium.org: Please review overall patch (as content/ and ui/ owner, you
seem uniquely qualified to review plumbing from one to the other).

msw@chromium.org: Please review changes in chrome/browser/ui/

erg@chromium.org: Please review ui/views/widget/desktop_aura/*x11*

Thanks all!

[Avi (use Gerrit), 2015-05-18 14:59:22]

I'm OK with the idea, but turning this into two code paths is rather
unfortunate.

I'm doing code searches for the callers of the Activate* functions and the
numbers are in the low dozens. Can we break this CL into several pieces and do
it better?

https://codereview.chromium.org/1135693006/diff/1/content/public/browser/web_...
File content/public/browser/web_contents_delegate.h (right):

https://codereview.chromium.org/1135693006/diff/1/content/public/browser/web_...
content/public/browser/web_contents_delegate.h:110: virtual void
ActivateContentsForUserGesture(WebContents* contents);
There's a lot of duplicate code in some of these functions. I know that
Activate* calls are all over the place, so it would be hard to add a bool
parameter |user_gesture| rather than doing it this way, but it's a bit
unfortunate.

[Elliot Glaysher, 2015-05-18 17:38:03]

The actual x11 message changes lg, but I agree with avi@ about the duplicate
interface is suboptimal.

[johnme, 2015-05-22 16:42:28]

On 2015/05/18 14:59:22, Avi wrote:
> I'm OK with the idea, but turning this into two code paths is rather
> unfortunate.
> 
> I'm doing code searches for the callers of the Activate* functions and the
> numbers are in the low dozens. Can we break this CL into several pieces and do
> it better?
On 2015/05/18 17:38:03, Elliot Glaysher wrote:
> The actual x11 message changes lg, but I agree with avi@ about the duplicate
> interface is suboptimal.

Ok, I've gone with adding a |bool user_gesture| to the Activate functions. I've
split this up into a chain of CLs, as follows:

- https://codereview.chromium.org/1153793003 adds user_gesture to
WebContentsImpl::Activate
- https://codereview.chromium.org/1153813003 adds user_gesture to
WebContentsDelegate::ActivateContents
- https://codereview.chromium.org/1149263003 adds user_gesture to
BaseWindow::Activate
- https://codereview.chromium.org/1158523002 adds user_gesture to
BaseWindow::Show

Each patch depends on the previous patch (since the later patches use the
user_gesture parameter provided by earlier patches).

avi: Can you check that this seems reasonable? If so, I'll go ahead and do the
next patch which will update Widget::Activate in similar fashion, and finally a
patch that updates x11_desktop_handler and friends. Thanks!