Work around --profiler=perf failure with recent linux perf.

tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py

Index: tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py
diff --git a/tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py b/tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py
index e1fe00e6aa4aae7e26eb2b92eb311bab970bc901..6d969231433591b9cb5221f35bfed967a3e308e8 100644
--- a/tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py
+++ b/tools/telemetry/telemetry/core/platform/profiler/perf_profiler.py
@@ -41,8 +41,10 @@ class _SingleProcessPerfProfiler(object):
           'mkdir -p ' + os.path.dirname(self._device_output_file))
     else:
       cmd_prefix = ['perf']
+    # In perf 3.13 --call-graph requires an argument, so use
+    # the -g short-hand which does not.
     self._proc = subprocess.Popen(cmd_prefix +
-        ['record', '--call-graph',
+        ['record', '-g',
          '--pid', str(pid), '--output', output_file],
         stdout=self._tmp_output_file, stderr=subprocess.STDOUT)