Read README.rst as UTF-8 regardless of external locale settings.

The sanitized environment used by Debian package build daemons forces
use of the barebones ‘POSIX’ locale.  Some versions of Python 3 will
refuse to read an UTF-8 file as text when this locale is in use,
unless explicitly told to use UTF-8 anyway.  Python 2 doesn’t support
the encoding= argument to open(), so to get setup.py to do the Right
Thing in a universal fashion, read the file as *binary* and then
explicitly decode it.

Author: zackw

setup.py

@@ -17,13 +17,16 @@ def run(self):
         self.run_command("build_ext")
         return _build_py.run(self)
 
+# Read a file that may or may not exist, and decode its contents as
+# UTF-8, regardless of external locale settings.
 def read_file(filename):
     filepath = os.path.join(
         os.path.dirname(os.path.dirname(__file__)), filename)
-    if os.path.exists(filepath):
-        return open(filepath).read()
-    else:
+    try:
+        raw = open(filepath, 'rb').read()
+    except OSError:
         return ''
+    return raw.decode('utf-8')
 
 def mecab_config(arg):
     return os.popen("mecab-config " + arg).readlines()[0].split()