MacViews: Fix WidgetTest.WidgetDeleted_InOnMousePressed

MacViews: Fix WidgetTest.WidgetDeleted_InOnMousePressed

WidgetTest.WidgetDeleted_InOnMousePressed was failing because
the test widget is initialized with a 0x0+0+0 bounds, which is then
resized with SetSize(). Since it's also a popup type, widget.cc does
not call either of SetInitialBounds{,ForFramelessWindow}. This resulted in
a window that was offscreen, so mouse events were not sent to it.

To fix, widgets with no initial bounds are put at the top left of the screen
rather than the bottom-left.

The test was also flaky because the event generator was not retaining
the NSWindow. The tests are specifically to cover the widget being destroyed.
Cocoa retains the window during event dispatch, so the event simulator
should as well.

BUG=378134
Committed: https://crrev.com/4e3993692cad3a7835c9cb47954f60f97fdf4c23
Cr-Commit-Position: refs/heads/master@{#319532}

[tapted, 2015-03-06 12:03:41]

tapted@chromium.org changed reviewers:
+ andresantoso@chromium.org, sadrul@chromium.org

[tapted, 2015-03-06 12:03:42]

Hi Andre and sadrul - please take a look (sadrul for ui/views/test owners)

Note this is split out from something more ambitious in
http://crrev.com/976963002 that tries to fix some other tests:
CloseWidgetDuringMouse{Press, Released}. It's related but turned out to be a
very different problem. Just disabling those tests might make more sense for
mac. But I bring it up, because the fix there might be an alternative to the
`retain` approach here to prevent flakiness. *but* I like this better :)

For Aura, the key difference is WidgetDeleted_InOnMousePressed tests
NativeWidgetAuras (parented on the root window which is never closed), and the
CloseWidgetDuringMouse.* tests test DestktopNativeWidgetAuras which lose their
dispatcher when the window is closed. But on Mac, the dispatcher is the
NSWindow, which Cocoa retains, so doesn't have the same issues (and for those
tests, the EventGenerator turns out having to simulate a lot in
http://crrev.com/976963002 -- probably too much).

[Andre, 2015-03-06 17:52:02]

LGTM

[sadrul, 2015-03-06 22:05:21]

lgtm

[tapted, 2015-03-06 22:49:50]

The CQ bit was checked by tapted@chromium.org

[commit-bot: I haz the power, 2015-03-06 22:51:03]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/987733002/1

[commit-bot: I haz the power, 2015-03-07 00:14:22]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-03-07 00:15:00]

Patchset 1 (id:??) landed as
https://crrev.com/4e3993692cad3a7835c9cb47954f60f97fdf4c23
Cr-Commit-Position: refs/heads/master@{#319532}