Make Frame's ScriptController an OwnPtr and remove the #include

Make Frame's ScriptController an OwnPtr and remove the #include

Merges http://trac.webkit.org/changeset/148373 by AwesomeWeinig
(but the projects have already diverged enough that the actual diff
looks pretty different).

Frame is included in almost 1000 files, so keeping it small speeds up
compilation.

This exposes a null-pointer deref bug in XMLHttpRequest that
happened to work until now. To fix, also merge Sam's
http://trac.webkit.org/changeset/148380:
"Check that the frame is not null (as it can be in cases like
http/tests/xmlhttprequest/detaching-frame-2.html).
We used to be getting lucky, in that shouldBypassMainWorldContentSecurityPolicy(),
the function that is ultimately called, only operates on global state. Now that we need
to actually dereference the Frame to get the ScriptController, we see this crash."

BUG=none

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=148570

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=148609

[Nico, 2013-04-17 04:27:19]

r?

[abarth-chromium, 2013-04-17 04:32:06]

What's the benefit of making this change?

[Nico, 2013-04-17 04:36:17]

Frame.h is included in almost 1000 files (
https://code.google.com/p/chromium/codesearch#search/&q=%23include%5C%20%5C%2...
), so making it smaller speeds up compilation. (Added to CL description.)

Also, touching ScriptController.h will now require less recompilation.

[abarth-chromium, 2013-04-17 04:39:53]

Ok.  Ideally we'd know how much faster it's making compilation, but LGTM even
without that.

[eseidel, 2013-04-17 06:20:01]

wat?  Wenig objected last time I tried this... or so I remember.  I must just be
getting old and senile.

[eseidel, 2013-04-17 06:20:26]

lgtm too.  I think this is worth doing even if it doesn't speed compiles.  It
will speed incremental rebuilds!

[Nico, 2013-04-17 19:30:01]

Committed patchset #5 manually as r148570 (presubmit successful).

[Nico, 2013-04-17 23:51:03]

This got reverted since it made a few XHR tests fail. The difference from patch
set 6 to 7 contains the fix (also merged from webkit, where the same problem
occurred). I verified that the failing tests now pass locally will reland.

[commit-bot: I haz the power, 2013-04-17 23:55:05]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/thakis@chromium.org/14323004/18002

[commit-bot: I haz the power, 2013-04-18 02:39:37]

Retried try job too often on linux_layout_rel for step(s) webkit_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=linux_layo...

[Nico, 2013-04-18 03:17:16]

Committed patchset #7 manually as r148609 (presubmit successful).