lib/ldb: Implement a bytes derived object for attributes py2/py3

lib/ldb: Implement a bytes derived object for attributes py2/py3

ldb attributes are either bytes (py3) or str (py2)

Some places in the code do str(res[0]['attribute'][0])
which results in
   'result' (py2)
  b'result' (py3)

or more commonly the attribute is used to construct a string e.g.
   "blah=" + res[0]['attribute'][0] + ",foo,bar=...."

giving
   "blah=result,foo,bar=...." (py2)
and very unhelpfully
   "blah=b'result',foo,bar=...." (py3)

lots of code already constructs various strings for passing to other
api using the above. To avoid many excessive
    res[0]['attribute'][0].decode('utf8')

code like 'res[0]['attribute'][0]'

will now return LdbBytes (a new object subclassing 'bytes') in py3
instead of bytes. This object has a custom '__str__' method which
attempts to return a string decoded to uft8. In Py2 this will behave as
it did previously (this is the safer option at the moment)

Signed-off-by: Noel Power <noel.power@suse.com>
Reviewed-by: Andreas Schneider <asn@samba.org>
Reviewed-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>