This caused some issues with printing nested objects that contain unicode, for example printing dictions now yields something like {u'ok': True, u'msg': u''}.

Summary of test changes:
    - Changed some doctests to test for dict equality instead of printing the dicts
        - there will be more of these...
    - Changed some pytest `with raises(Error, match='')` to use a regexp
        - there will be more of these...
    - Changed debug log test to include a .format(u=UNICODE_PREFIX) bit that conditionally adds a u-prefix in Python 2. This was simpler than changing the match to a regexp and escaping a whole bunch o stuff.

This also unifies very preliminary error checking on student_input

Atom's python-tools find-usage command seems not to miss instances of str wrapped in parens this way...

This is handled by ListGrader.ensure_text_inputs now, and tested by `tests/test_listgrader.py` function `test_errors`

it's better =)