Add the local blink_perf benchmark that supports running any subdir of WebKit/PerformanceTest

tools/perf/contrib/blink_perf_cmdline/blink_perf_cmdline.py

Index: tools/perf/contrib/blink_perf_cmdline/blink_perf_cmdline.py
diff --git a/tools/perf/contrib/blink_perf_cmdline/blink_perf_cmdline.py b/tools/perf/contrib/blink_perf_cmdline/blink_perf_cmdline.py
new file mode 100644
index 0000000000000000000000000000000000000000..1e406f34c599d89fd9dbce7a5cb303f10732578a
--- /dev/null
+++ b/tools/perf/contrib/blink_perf_cmdline/blink_perf_cmdline.py
@@ -0,0 +1,26 @@
+# Copyright 2017 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+from benchmarks import blink_perf
+
+
+# pylint: disable=protected-access
+class BlinkPerfAll(blink_perf._BlinkPerfBenchmark):
+
+  @classmethod
+  def Name(cls):
+    return 'blink_perf'
+
+  @classmethod
+  def ProcessCommandLineArgs(cls, parser, args):
+    if len(args.positional_args) > 2:
+      parser.error('Too many arguments.')

[Xianzhu, 2017/05/10 02:55:37]

1. Does it still support options like --browser, --show-stdout?
2. Can the user get instructions with --help?

[nednguyen, 2017/05/10 17:23:42]

1. Yes.
2. Good question. It could be hard to augment this benchmark's own help related
to positional args to the overall help, but lemme see what I can do here.

[Xianzhu, 2017/05/10 18:01:28]

Like charliea asked in https://codereview.chromium.org/2873183002/, I also
prefer options to positional args. Options requires more typing, but provides
better way for users to discover features.

[nednguyen, 2017/05/10 20:30:44]

Done. There is no need for the other CL then.

+
+  def CreateStorySet(self, options):
+    if len(options.positional_args) > 1:
+      path = options.positional_args[1]

[Xianzhu, 2017/05/10 02:55:37]

I think it would be more convenient if the path is relative to
BLINK_PERF_BASE_DIR.

[nednguyen, 2017/05/10 17:23:42]

whynotboth.gif? since the other way is good for commandline auto-complete.

[Xianzhu, 2017/05/10 18:01:28]

Supporting both will be great :), just like run-webkit-tests and 'content_shell
--run-layout-test' supporting both real paths and paths relative to LayoutTests
directory. Can you add it in this CL?

[nednguyen, 2017/05/10 20:30:44]

Done.

+    else:
+      path = blink_perf.BLINK_PERF_BASE_DIR
+      print 'Running all blink_perf tests. This will take a long time...'
+    return blink_perf.CreateStorySetFromPath(path, blink_perf.SKIPPED_FILE)