Fix import of certain VII stems

[eddieantonio]

This has been fixed in the source code in a very ad hoc way. Basically, there was an error in the export of the source that has now been corrected. I've gone into src/fst/stems/verb_stems.lexc and changed all the errant VIIn that were classified as VIIw. This should fix these issues for now, but we will have to double check that no -n finals are in VIIw and vice versa when we re-export from source to our FST in the next update.

Originally posted by @atticusha in UAlbertaALTLab/morphodict#573 (comment)

[aarppe]

To be explicitly clear, the problem was a miscoding in the source database, not in our script turning this database into the LEXC code.

The one challenge as to why I'm wary of rerunning the LEXC generation script on an updated version of the database is that we still need to add manually the fst-stems which are not in the source database, and some regressions tests are needed as well. While we can programmatically add the FST stems to the current database, these FST stems need to be added to every new entry in the database as well.