The code previously converted an unicode object into unicode, but in Python 2.X
models actually have to specify 'str' (GTK expects a gchararray). So encode to
UTF-8 instead to get what GTK expects.

Add corresponding test case.