Add typing, switch to list comprehensions where possible,
encapsulate all methods within new chunker implementation,
use dataclass instead of unmanged dictionary,
list dependencies in setup installation line.
Fix token counting bug due to static initialization of
`semchunk.Chunker`.
Use expanded chunk typing (from -core) including
embedding-specific and gen-specific texts.
Signed-off-by: Panos Vagenas <35837085+vagenas@users.noreply.github.com>
