Properly recreate swapped out RenderView.

Properly recreate swapped out RenderView.

This CL fixes how RenderView is recreated after a process crash. Due to not
recreating the RenderFrameProxy in the case the RenderView is swapped out, the
WebView ends up with no mainFrame() and crashes in various ways.

This is a reland of https://codereview.chromium.org/1408743005/ with the
refactoring portion committed separately in https://codereview.chromium.org/1415303002/.

BUG=357747, 544271
Committed: https://crrev.com/2fd02526ca100834b7a2b13304203244670d7bb3
Cr-Commit-Position: refs/heads/master@{#355802}

[nasko, 2015-10-22 22:54:26]

Description was changed from

==========
Properly recreate swapped out RenderView.

This CL fixes how RenderView is recreated after a process crash. Due to not
recreating the RenderFrameProxy in the case the RenderView is swapped out, the
WebView ends up with no mainFrame() and crashes in various ways.

BUG=357747, 544271
==========

to

==========
Properly recreate swapped out RenderView.

This CL fixes how RenderView is recreated after a process crash. Due to not
recreating the RenderFrameProxy in the case the RenderView is swapped out, the
WebView ends up with no mainFrame() and crashes in various ways.

This is a reland of https://codereview.chromium.org/1408743005/ with the 
refactoring portion committed separately in
https://codereview.chromium.org/1415303002/.

BUG=357747, 544271
==========

[nasko, 2015-10-22 23:48:09]

nasko@chromium.org changed reviewers:
+ creis@chromium.org

[nasko, 2015-10-22 23:48:10]

Hey Charlie,
This is a reland of a CL you already reviewed. It should be simple enough and
I've included this time a comment on the actual fix part for this bug.
Thanks in advance!
Nasko

https://codereview.chromium.org/1412173006/diff/20001/content/browser/frame_h...
File content/browser/frame_host/render_frame_host_manager.cc (left):

https://codereview.chromium.org/1412173006/diff/20001/content/browser/frame_h...
content/browser/frame_host/render_frame_host_manager.cc:1805: proxy = nullptr;
This is the main part of this fix. When a proxy exists, it must be recreated on
the renderer side. Not nulling the pointer here will allow this to happen later
in this method as part of InitRenderView.

[Charlie Reis, 2015-10-22 23:54:15]

LGTM.

[nasko, 2015-10-23 15:16:49]

The CQ bit was checked by nasko@chromium.org

[commit-bot: I haz the power, 2015-10-23 15:17:06]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1412173006/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1412173006/20001

[commit-bot: I haz the power, 2015-10-23 16:41:51]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-10-23 16:42:35]

Patchset 2 (id:??) landed as
https://crrev.com/2fd02526ca100834b7a2b13304203244670d7bb3
Cr-Commit-Position: refs/heads/master@{#355802}