Fix some more memory leaks.
Call error_ret() in decode_str().  It was called in some other places,
but seemed inconsistent.  It is safe to call PyTokenizer_Free() after
calling error_ret().
1 file changed