And add tests when reading non-ASCII strings and comments (both from Python strings and from files). Fixes: #10
Previously it was not tested, and broke on Python 2 `unicode` inputs.