Convert //net and //chromecast to std::unordered_*

chromecast/base/device_capabilities_impl.cc

Index: chromecast/base/device_capabilities_impl.cc
diff --git a/chromecast/base/device_capabilities_impl.cc b/chromecast/base/device_capabilities_impl.cc
index c3dfea0d07e590eb5e30957f536874a0e3fd1d0a..60c2153ccf632ea56960a206b4b8316b837a6f5c 100644
--- a/chromecast/base/device_capabilities_impl.cc
+++ b/chromecast/base/device_capabilities_impl.cc
@@ -149,11 +149,9 @@ void DeviceCapabilitiesImpl::Register(const std::string& key,
   DCHECK(validator);
 
   base::AutoLock auto_lock(validation_lock_);
-  bool added =
-      validator_map_.add(key, base::WrapUnique(new ValidatorInfo(validator)))
-          .second;
   // Check that a validator has not already been registered for this key
-  DCHECK(added);
+  DCHECK_EQ(0u, validator_map_.count(key));
+  validator_map_[key] = base::WrapUnique(new ValidatorInfo(validator));

[davidben, 2016/04/11 19:27:58]

This technically changes behavior in the failed DCHECK case. If we want to
completely emulate the old behavior, it'd be:

  bool added = validator_map_.insert(std::make_pair(key, base::WrapUnique(new
ValidatorInfo(validator)))).second;
  DCHECK(added);

But that's a mouthful and there's a DCHECK, so I assumed this version is
preferable.

[lcwu1, 2016/04/11 21:43:03]

Yes, I like the new implementation better. Thanks for doing this.

lgtm

 }
 
 void DeviceCapabilitiesImpl::Unregister(const std::string& key,