Remove hit test from dispatchDecidePolicyForNavigationAction

Remove hit test from dispatchDecidePolicyForNavigationAction

There's no need to do a hit test + layout in order to decide the policy for a
given navigation action. The originatingNode parameter is only used by a few
test codepaths that printf() information. This hit test forces a layout at an
awkward time which can be expensive and/or not do the expected thing.

BUG=232770
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=148586

[jamesr, 2013-04-17 20:50:13]

This appears to be old as time itself.  To actually remove the node will require
chromium-side changes.  I think the plan is to:

1.) Implement an override of decidePolicyForNavigation() without the WebNode
parameter in render_view_impl.cc, land it
2.) Change the signature in WebFrameClient.h to not have the WebNode, land it,
roll it
3.) Remove render_view_impl.cc's decidePolicyForNavigation(..., WebNode, ...)
implementation.

ditto for the test implementations - there's one in chromium and several in
Blink.

[Nate Chapin, 2013-04-17 20:53:11]

lgtm

[jamesr, 2013-04-17 21:36:06]

Turns out a few layout tests have the printf in the output:

fast/loader/policy-delegate-action-hit-test-zoomed.html appears to actually be
testing the hit testing logic (why????).  I think this should just be removed.

fast/encoding/mailto-always-utf-8.html has this expected output:
Policy delegate: attempt to load
mailto:?subject=%E3%82%BF%E3%82%A4%E3%83%88%E3%83%AB&body=%E6%9C%AC%E6%96%87
with navigation type 'link clicked' originating from #text > SPAN > A > BODY >
HTML > #document
Test encoding of mailto URLs. Click on the URL - a new message with Japanese
subject and body should open in your mail client.

Subject : Japanese Body : Japanese

I think the answer is to just delete the node path.  Other tests with the same
thing:

fast/forms/mailto/formenctype-attribute-button-html.html
fast/forms/mailto/formenctype-attribute-input-html.html
http/tests/misc/policy-delegate-called-twice.html
http/tests/misc/redirect-to-external-url.html
http/tests/security/feed-urls-from-remote.html

I'll rebaseline these

[jamesr, 2013-04-17 21:43:41]

Going through history it looks like Apple used the node for some policy thingy
or another but we never have.

[jamesr, 2013-04-17 22:35:47]

Committed patchset #2 manually as r148586 (presubmit successful).