Encode characters to utf-8 int the C string evaluator.

Our policy is not to deal with unicode() instances.

For some reason, this fixes a bug that showed up in the "OVM" app
bundle, but not the dev build running CPython.  There was a LookupError
saying the ascii codec wasn't found.

core/word_compile.py

@@ -70,7 +70,7 @@ def EvalCStringToken(id_, value):
  elif id_ in (Id.Char_Unicode4, Id.Char_Unicode8):
    s = value[2:]
    i = int(s, 16)
    return unichr(i)
    return unichr(i).encode('utf-8')  # Stay in the realm of bytes

  else:
    raise AssertionError