Clean up a few startup and shutdown dependencies which should fix some of the...

Clean up a few startup and shutdown dependencies which should fix some of the extension
browser_tests.  Specifically, this includes:
* make ExtensionsService own ExtensionProcessManager so that they have the same lifetime
* git rid of notifications from unload all to avoid treating these notifications like real unload / uninstall events
* fix teardown logic in ExtensionShelfModelTest
* ExtensionView removes itself from the view hierarchy on destruction
* ExtensionProcessManager waits for EXTENSIONS_READY to load background pages

BUG=15080
TEST=browser_tests.exe --gtest_filter=ExtensionShelfModelTest


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=19122

[Matt Perry, 2009-06-23 23:05:35]

LGTM

http://codereview.chromium.org/147051/diff/1/3
File chrome/browser/extensions/extension_shelf_model_unittest.cc (right):

http://codereview.chromium.org/147051/diff/1/3#newcode108
Line 108: // before the rest of InProcessBrowserTest.
update comment to explain why this isn't in TearDown

[rafaelw, 2009-06-24 23:04:21]

http://codereview.chromium.org/147051/diff/1/6
File chrome/browser/extensions/extension_process_manager.cc (right):

http://codereview.chromium.org/147051/diff/1/6#newcode36
Line 36: if (service && service->is_ready())
Can this test ever return true?

[Erik does not do reviews, 2009-06-25 16:19:08]

http://codereview.chromium.org/147051/diff/1/6
File chrome/browser/extensions/extension_process_manager.cc (right):

http://codereview.chromium.org/147051/diff/1/6#newcode36
Line 36: if (service && service->is_ready())
On 2009/06/24 23:04:22, rafaelw wrote:
> Can this test ever return true?

Good point.  It can't any more.  I'll fix this in a follow-on.