Add bisect-config-trying script and a few example configs.

tools/auto_bisect/configs/android.bisect.blink_perf.cfg

Index: tools/auto_bisect/configs/android.bisect.blink_perf.cfg
diff --git a/tools/auto_bisect/configs/android.bisect.blink_perf.cfg b/tools/auto_bisect/configs/android.bisect.blink_perf.cfg
new file mode 100644
index 0000000000000000000000000000000000000000..b7af657ca79338588f70eabcff018905b98caf29
--- /dev/null
+++ b/tools/auto_bisect/configs/android.bisect.blink_perf.cfg
@@ -0,0 +1,19 @@
+# This config should find the regression of http://crbug.com/423495.
+# It should reproduce a > 10% change with the culprit CL being a blink change.
+# http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus7_perf_bisect/builds/394
+
+config = {
+  'command': 'tools/perf/run_benchmark -v --browser=android-chrome-shell blink_perf.layout',
+  'good_revision': '297772',
+  'bad_revision': '297779',
+  'metric': 'character_fallback/character_fallback',
+  'repeat_count': '10',

[prasadv, 2014/10/21 23:44:47]

I would prefer reducing repeat count if that gives the same results.

[qyearsley, 2014/10/23 02:51:01]

Right -- the less time the better. Although, it'd be good if the config reliably
reproduced the regression, which it might not if the repeat count is 1.

+  'max_time_minutes': '10',
+  'truncate_percent': '25',
+  'bug_id': '423495',
+  'gs_bucket': 'chrome-perf',
+  'builder_host': 'master4.golo.chromium.org',

[prasadv, 2014/10/21 23:44:47]

Not related to this CL, but I think we should get rid of builder_host and
builder_port from config.

[qyearsley, 2014/10/23 02:51:01]

Agreed. The CL that does that should also update these configs.

+  'builder_port': '8341',
+}
+
+# Workaround git try issue, see crbug.com/257689