encrypt_password and decrypt_password were previously expecting to

work with byte strings, but given the nature of Python 2, those were

implicitly being converted elsewhere to Unicode strings. In Python 3,

they're not so interchangeable, and requiring/returning byte strings

makes for too much hassle, breaking all sorts of code.
These functions now expect to be dealing with Unicode strings. Both

return Unicode strings and accept either byte strings or Unicode strings

as values. This is actually safer than what we had before, since they

were always turning into Unicode at some point, and now it's at a level

that we control.

All unit tests pass on Python 2.7.
Unit tests that can be run on Python 3.x that used to break due to

password string types now pass.
Manually tested saving and loading passwords for repositories.