Because CSV has no way of indicating a content character set, writing the
file with a Unicode byte-order mark is useful to declare the content in
UTF-8 encoding.