Move chrome_proxy tests from under tools/perf to tools/chrome_proxy/integration_tests.

tools/perf/run_tests

Index: tools/perf/run_tests
diff --git a/tools/perf/run_tests b/tools/perf/run_tests
index ed1429cb7f9b5c84b51df529558b6d9010d1ab89..3cc6b94c74a7cd2faa15bf4f71cb19a43aeccdef 100755
--- a/tools/perf/run_tests
+++ b/tools/perf/run_tests
@@ -19,4 +19,14 @@ from telemetry.unittest import run_tests
 if __name__ == '__main__':
   base_dir = os.path.dirname(os.path.realpath(__file__))
   run_tests.environment = run_tests.Environment(base_dir, [base_dir])
-  sys.exit(run_tests.RunTestsCommand.main())
+  test_result = run_tests.RunTestsCommand.main()
+
+  # Continue with chrome_proxy tests which are under a different directory.
+  chrome_proxy_dir = os.path.realpath(
+      os.path.join(os.path.dirname(__file__), os.pardir, 'chrome_proxy'))
+  sys.path.append(chrome_proxy_dir)
+  run_tests.environment = run_tests.Environment(
+      chrome_proxy_dir, [chrome_proxy_dir])
+
+  test_result += run_tests.RunTestsCommand.main()
+  sys.exit(test_result)

[tonyg, 2014/07/17 18:27:52]

If the benchmark is happy, then I'm happy with this.

But I know Dave has a lot of plans here, so Dave would you mind double checking
this?

[dtu, 2014/07/17 22:17:38]

I think I'd prefer another run_tests or run_unittests in tools/chrome_proxy/.
There're a few points of concern here.
- Not sure if the buildbot log parser will work with multiple separate gtest
outputs in the same stream.
- Telemetry zip archive doesn't contain chrome_proxy/, run_tests will fail when
running from the archive.
- By design we disallow multiple top_level_dirs. nduca will know more. Also we
disallow this cross-directory dependency.