I'm not sure how useful in practice this is, as we don't offer a codec
configuration default for it, but I'm not sure you'd never do it, and the
existing code was in the wrong order.

Wire format data is now done via a dns.wire.Parser, which does all of the
bookkeeping and also provides convenience routines (e.g. get_uint16() or
get_name()).

Classes inherit from object by default; there's no need to explicitly
include this.

Replace super(Foo, self) with super().

This does two things.

1) Makes the file parameter optional, removing the requirement that
callers create an io.BytesIO object if all they want is the bytes.  This
is done by renaming all of the rdata subclass to_wire() methods to
_to_wire(), and making dns.rdata.to_wire() create the io.BytesIO
internally if needed, and then delegate to the subclass.

2) Add canonicalize=False parameter, allowing code reuse for conversion
to wire format and conversion to hashable format.  This also removes all
of the rdata subclass to_digestable routines that are no longer needed,
as dns.rdata.to_digestable will work for all rdata types.

There's no need to truncate the result of __hash__(); the runtime will
already truncate it to sys.hash_info.width.