PyUnicode_EncodeUTF8() already returns the created object with the
proper reference count.  This fixes an Insure reported memory leak.