[Snowball-discuss] Minor bug in utf-8 handling

[Olly Betts]

I think I've spotted a bug in the handling of 3 byte utf-8 sequences
while reading the code.  Both get_utf8 and get_b_utf8 fetch the third
byte with *p when they should use p[c]:

http://oligarchy.co.uk/xapian/patches/snowball-3byte-utf8-bugfix.patch

In current stemmers, this is probably harmless, as the characters in use
in the languages snowball has stemmers for encode as one or two byte
utf-8 sequences.

I also improved the comment before skip_utf8.

Cheers,
    Olly