tests: avoid implicit conversion of str to unicode
On Python 2, str.encode('utf-8') implicitly converts the string to unicode and
then back to str.
Using _sys2bytes() ensures that opt is only encoded on Python 3, where opt is
unicode.
Although contrived, before this change, a UnicodeDecodeError could be triggered
on Python 2 when passing non-ascii values to --extra-config-opt.
