One thing I haven't done yet is update the types db.
This is just equating char-sets and isets - if we want
the two disjoint I can make a char-sets a thin wrapper
around isets.
Alternately we can drop iset and just define everything
in terms of chars to begin with, but I think iset is useful.
The libchicken size on my machine is about 100k larger.
This should be reclaimed after the next change which
would redefine the irregex char-sets in terms of SRFI-14.
The change after that defines the utf8 egg char-sets in
terms of the core.
The char-sets also take less space at runtime. There
are 17 default char-sets, each of which take 256 bytes
in the old SRFI-14, and are guaranteed to be smaller in
the new representation (usually only 5 words).