benchmarks: fix run_smoke_bench after upgrading to python3

Need to decode result of `subprocess.check_output`

Author: eeckstein

benchmark/scripts/run_smoke_bench

@@ -307,11 +307,12 @@ def report_code_size(opt_level, old_dir, new_dir, architecture, platform, output
 
 
 def get_codesize(filename):
-    output = subprocess.check_output(["size", filename]).splitlines()
-    header_line = output[0]
-    data_line = output[1]
+    output = subprocess.check_output(["size", filename])
+    lines = output.decode('utf-8').splitlines()
+    header_line = lines[0]
+    data_line = lines[1]
     if header_line.find("__TEXT") != 0:
-        sys.exit("unexpected output from size command:\n" + output)
+        sys.exit("unexpected output from size command:\n" + lines)
     return int(data_line.split("\t")[0])
 
 
@@ -361,7 +362,8 @@ performance team (@eeckstein).
   <summary><strong>Hardware Overview</strong></summary>
 
 """
-    po = subprocess.check_output(["system_profiler", "SPHardwareDataType"])
+    output = subprocess.check_output(["system_profiler", "SPHardwareDataType"])
+    po = output.decode('utf-8')
     for line in po.splitlines():
         selection = [
             "Model Name",