In contrast to the preview CSV reader, the actual ``ImportModel`` did not respect any encoding values provided. This leads to bugs with any non utf8-compliant characters. This PR fixes the problem by adding the appropriate encoding filter (copied from the preview reader).