gpo: Fix line length calculation for byte strings (bug 1598)

gpo incorrectly calculates the length of podcasts containing non-ASCII
characters (e.g. Russian ones), which results in the "[DONE]" label
being closer to the message and not aligned with other "[DONE]" labels.
This patch fixes it by converting byte strings to unicode before
calculating the string length, resulting in properly aligned output.

bin/gpo

@@ -78,6 +78,8 @@ if os.path.exists(src_dir) and os.path.exists(data_dir) and \
 import gpodder
 _ = gpodder.gettext
 
+from gpodder import util
+
 # Use only the gPodder API here, so this serves both as an example
 # and as a motivation to provide all functionality in the API :)
 from gpodder import api
@@ -99,7 +101,7 @@ class gPodderCli(object):
         self._current_action = ''
 
     def _start_action(self, msg, *args):
-        line = msg % args
+        line = util.convert_bytes(msg % args)
         if len(line) > self.COLUMNS-7:
             line = line[:self.COLUMNS-7-3] + '...'
         else: