enhancement to use column numbers rather than column names

[simbo1905]

If there are files that don't have a column header it would help to be able to specify the key and ignored columns by column number e.g., 0,3,6. This can be simulated by a wrapper script that adds out a top line 0,1,2,3,... into the two files. It would be helpful if cvsdiff supported this directly to avoid having to insert header lines into both files.