fix preprocess byte encoding on python 3

[chanind]

This PR fixes a bug in the preprocess step for Python 3, which is writing tokens as b'token' instead of just token, as pointed out in @appleternity in https://github.com/swabhs/open-sesame/issues/61#issuecomment-789312514. Python 2 is officially dead as of Jan 1, 2020, so it should be a priority to make sure that this code works on Python 3 rather than 2.