Enable features API for content shell and webview

content/shell/browser/shell_browser_main_parts.cc

Index: content/shell/browser/shell_browser_main_parts.cc
diff --git a/content/shell/browser/shell_browser_main_parts.cc b/content/shell/browser/shell_browser_main_parts.cc
index 9b17cd58040a09ad74ba1528ea318e995e140ca1..5b5d7408cb9a99ff0e8b19286e6545b314c7d2f2 100644
--- a/content/shell/browser/shell_browser_main_parts.cc
+++ b/content/shell/browser/shell_browser_main_parts.cc
@@ -7,6 +7,7 @@
 #include "base/base_switches.h"
 #include "base/bind.h"
 #include "base/command_line.h"
+#include "base/feature_list.h"
 #include "base/files/file_path.h"
 #include "base/files/file_util.h"
 #include "base/message_loop/message_loop.h"
@@ -133,6 +134,23 @@ void ShellBrowserMainParts::PreEarlyInitialization() {
 #endif
 }
 
+int ShellBrowserMainParts::PreCreateThreads() {
+  // The only case GetInstance() returns null should be when you are running a
+  // C++ test, in which case currently we do not support command line features.
+  // This allows Test::SetUp() to enable/disable features as necessary. See
+  // crbug.com/596021 for more details.
+  if (!base::FeatureList::GetInstance()) {
+    const base::CommandLine* command_line =
+        base::CommandLine::ForCurrentProcess();
+    scoped_ptr<base::FeatureList> feature_list(new base::FeatureList);
+    feature_list->InitializeFromCommandLine(
+        command_line->GetSwitchValueASCII(switches::kEnableFeatures),
+        command_line->GetSwitchValueASCII(switches::kDisableFeatures));
+    base::FeatureList::SetInstance(std::move(feature_list));

[Alexei Svitkine (slow), 2016/03/24 17:16:37]

Instead of requiring every BrowserMainParts implementation to do this, can this
be done in a place where it's only done once?

I think the right place is the current call to
base::FeatureList::InitializeInstance() from browser_main_loop.cc - is there a
reason we simply can't add the enable_features & disable_features params to that
function?

[Changwan Ryu, 2016/03/28 07:09:27]

Makes perfect sense that this should be done inside browser_main_loop.cc, so
I've moved it there.

Extending InitializeInstance() with the two parameters seemed a bit unclear to
me as to whether it would override the existing feature list, or newly
instantiate one, and in order to avoid string searches in the case of chrome and
tests, I left the code as is. Please let me know if you have a better idea.

[Alexei Svitkine (slow), 2016/03/29 15:08:03]

I did not mean that the two params should override an existing feature list if
it exists - only register it with those params if there isn't one already. I
understand doing so will be a little inefficient since you have to look up the
two command line params before the instance check is done, but I think that's a
trivial performance cost especially since this is only done once on start up.

+  }
+  return 0;
+}
+
 void ShellBrowserMainParts::InitializeBrowserContexts() {
   set_browser_context(new ShellBrowserContext(false, net_log_.get()));
   set_off_the_record_browser_context(