tests/test.py: Make script compatible with python3

Currently the test script assumes the returned web page is always encoded in UTF-8, which might not be the case. It is likely that the real encoding is `ISO-8859-1` and that running the script under python3 will raise error:

```
< 
{ [11931 bytes data]
100 11923    0 11923    0     0   184k      0 --:--:-- --:--:-- --:--:--  187k
Traceback (most recent call last):
  File "tests/test.py", line 106, in <module>
    line = str(line, 'utf8')
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa0 in position 1105: invalid start byte
```
This pull request makes no assumption on the returned data and send the raw output to stderr as `bytes`. This should allow the test script to finish successfully when `tests/test.py` is run by python3.

tests/test.py

@@ -103,8 +103,10 @@ try:
                 if stage == 4 and fd == p4.stdout:
                     stage = 5
             if bytes != str:
-                line = str(line, 'utf8')
-            sys.stderr.write(line)
+                line = bytes(line)
+                sys.stderr.buffer.write(line)
+            else:
+                sys.stderr.write(line)
 
         if stage == 3 and p3 is not None:
             fdset.remove(p3.stdout)