Added checking for UnicodeErrors when reading in the preprint-report numbers and the periodical titles knowledge bases: if a line with badly encoded characters is read in, refextract will die with an error; Docstring line-length shortening after pylint convention complaints;