Avoid rebuilding remoting host every time gyp runs

remoting/remoting.gyp

Index: remoting/remoting.gyp
diff --git a/remoting/remoting.gyp b/remoting/remoting.gyp
index 713ee4c3e00314f5a6c558d3600fb0e117d522cf..20b7eec8162ba6f1d0fb06cb7619d81dd487eb7d 100644
--- a/remoting/remoting.gyp
+++ b/remoting/remoting.gyp
@@ -59,16 +59,17 @@
       ['OS=="win"', {
         'host_plugin_extension': 'dll',
         'host_plugin_prefix': '',
-        # Use auto-generated CLSIDs to make sure that the newly installed COM
+
+        # Each CLSID is a hash of the corresponding interface ID and the
+        # current version string to make sure that the newly installed COM

[Jamie, 2014/02/05 18:53:58]

How will this impact dev builds on a machine? I've lost many hours debugging
code because it turns out the version that was running was not the version I was
editing. It feels like this might increase the risk of that. Unfortunately, I
think that's fundamentally at odds with the thing we're trying to fix in this
CL, so I don't think there's a lot we can do; I just wanted to raise it as a
concern in case you have any suggestions.

[Sergey Ulanov, 2014/02/05 19:47:16]

I think potentially the IDs can be generated randomly as a separate build step. 
Problem is that the build target that generates IDs would need to depend on all 
code compiled into the host to make sure IDs are updated when code changes, but 
even then it wouldn't catch all cases. 

IMO proper solution is to stop using COM. FWIW we won't need COM for daemon
controller once NPAPI plugin is removed and I hope we can refactor
RDPDesktopSession to make it work without COM. So all COM dependencies can be
removed in the future, and then CLSID generation won't be a problem.

         # classes will be used during/after upgrade even if there are old
         # instances running already.
-        # The parameter at the end is ignored, but needed to make sure that the
-        # script will be invoked separately for each CLSID. Otherwise GYP will
-        # reuse the value returned by the first invocation of the script.
+        # The IDs are not random to avoid rebuilding host when it's not
+        # necessary.
         'daemon_controller_clsid':
-            '<!(python -c "import uuid; print uuid.uuid4()" 1)',
+            '<!(python -c "import uuid; print uuid.uuid5(uuid.UUID(\'655bd819-c08c-4b04-80c2-f160739ff6ef\'), \'<(version_full)\')")',
         'rdp_desktop_session_clsid':
-            '<!(python -c "import uuid; print uuid.uuid4()" 2)',
+            '<!(python -c "import uuid; print uuid.uuid5(uuid.UUID(\'6a7699f0-ee43-43e7-aa30-a6738f9bd470\'), \'<(version_full)\')")',

[Jamie, 2014/02/05 18:53:58]

I was originally concerned about duplicating these uuids, but I don't think it's
necessary for them to be the same as the corresponding interface ids, correct?
Maybe just use two random uuids here and update the comment so that we don't
have to worry about it in the future.

[Sergey Ulanov, 2014/02/05 19:47:16]

Correct. I've updated the comments to make it clear that the GUIDs don't have to
match the interfaces. Kept the actual values.

       }],
     ],
     'remoting_locales': [