[TimeZoneMonitor] Decouple renderer side impl from content to blink.

chrome/test/base/run_all_unittests.cc

Index: chrome/test/base/run_all_unittests.cc
diff --git a/chrome/test/base/run_all_unittests.cc b/chrome/test/base/run_all_unittests.cc
index fd8928d4023a1b69411a536f9bb65241bcd5b5ff..17d4e8c0f9b6c88dd1cb54164ef39ef604e26019 100644
--- a/chrome/test/base/run_all_unittests.cc
+++ b/chrome/test/base/run_all_unittests.cc
@@ -8,13 +8,11 @@
 #include "base/test/test_io_thread.h"
 #include "chrome/test/base/chrome_unit_test_suite.h"
 #include "content/public/test/unittest_test_suite.h"
-#include "mojo/edk/embedder/embedder.h"
 #include "mojo/edk/test/scoped_ipc_support.h"
 
 int main(int argc, char **argv) {
   content::UnitTestTestSuite test_suite(new ChromeUnitTestSuite(argc, argv));
 
-  mojo::edk::Init();

[blundell, 2016/10/26 18:08:05]

Are all the test changes around Mojo initialization still worth doing in this CL
given the decision around it being OK to not initialize Mojo in
blink::initialize()?

   base::TestIOThread test_io_thread(base::TestIOThread::kAutoStart);
   mojo::edk::test::ScopedIPCSupport ipc_support(test_io_thread.task_runner());