accent encoding / expected charcodes are wrong

[drzraf]

Most character are using the "\uXXXX" notation, but the values for XXXX are taken from the ISO version of the snowball files.
The FrenchStemmer r_un_accent function endup using things like:
if (!sbp.e_s_b(1, "\u00E9"))
instead if
if (!sbp.e_s_b(1, "\uC3A9"))

That's an issue for program like https://github.com/MihaiValentin/lunr-languages which intend to use this stemmer for client-side full-text search purposes.