In sed, 's/\n//' only works if you do an 'N' to read a second line into the buffer.

Once you have the table into something like a csv file, you need to sort and you may need to deal with excess " and such optional encoding. If there are embedded linefeeds in columns, you need to encode them so lines are rows. If the rows are unique, comm is a good tool, else either diff or preprocess sorted through uniq -c and sed to make dups into unique lines with a count suffix before comm.