fix handling of one-row data frames

add new argument 'col.names', with default TRUE. If
FALSE, column names of data frame are ignored

'row.names' may now be either logical (as before) or a
string, which is then used as column name