Improve unit testing of socket API (reland).

Improve unit testing of socket API (reland).

Add socket_api_unittest.cc, which calls Socket API functions directly.
This skips the whole browser_tests setup that is relatively slow and
expensive. The current test merely validates the socket-type argument
handling, but the next phase will simulate blocking socket operations.

Reland: I foolishly ran critical code inside a DCHECK, semi-thinking
that test code would run only in debug mode. This is hardly the case;
in fact, the ASAN bots run in release all the time.

BUG=112902
TEST=see above

Originally committed as 148856, then reverted in 148859. The only change
vs. http://codereview.chromium.org/10795052/ is that the DCHECK in
http://codereview.chromium.org/10795052/diff/27002/chrome/browser/extensions/test_extension_system.cc
is now a CHECK.

[miket_OOO, 2012-07-30 17:17:49]

asargent for entire CL, darin for chrome_tests.gypi OWNERS.

[Mihai Parparita -not on Chrome, 2012-07-30 17:36:16]

On Mon, Jul 30, 2012 at 10:17 AM, <miket@chromium.org> wrote:

> Reland: I foolishly ran critical code inside a DCHECK, semi-thinking
> that test code would run only in debug mode. This is hardly the case;
> in fact, the ASAN bots run in release all the time.
>

There are also release tryservers (not part of the default set), like
win_rel, linux_rel and mac_rel.

Mihai

[miket_OOO, 2012-07-30 17:48:39]

> There are also release tryservers (not part of the default set), like
> win_rel, linux_rel and mac_rel.
> 
> Mihai

The reason why the original CL made it past linux_rel
(http://build.chromium.org/p/tryserver.chromium/builders/linux_rel/builds/4303...
-- see that SocketUnitTest ran) is scary. The BrowserThread::ID enum UI is first
in the list and thus zero. The UM happened to be zero, too, so it all just
worked. It wasn't until a stricter compiler looked at it that it blew up.

[miket_OOO, 2012-07-30 17:59:11]

Reviewers: this CL just made it past the unit_tests step on
http://build.chromium.org/p/tryserver.chromium/builders/linux_asan/builds/491,
and I confirmed it's building with "--target Release." So I believe I've
identified and fixed the issue with the original CL.

[miket_OOO, 2012-07-30 18:05:39]

On 2012/07/30 17:59:11, miket wrote:
> Reviewers: this CL just made it past the unit_tests step on
> http://build.chromium.org/p/tryserver.chromium/builders/linux_asan/builds/491,
> and I confirmed it's building with "--target Release." So I believe I've
> identified and fixed the issue with the original CL.

I've closed this issue and reopened the original
http://codereview.chromium.org/10795052/. Thanks to Mihai for letting me know
this was possible.

[darin (slow to review), 2012-07-30 19:33:09]

LGTM again