- merging can now be done directly using knot_rrs_t

 - tests are passing, however some work is still pending, namely storage
   of additionals for freeing.
 - removed some obsolete node API functions

- if adding apex NS, adjust counter

- Do not merge incoming RRs
- Duplicated add/remove checks improved in ddns
- Remove stuff directly from RRSet, do not return

- removing and adding already existing record
- leak solved, but need to move old a new data lists from
  zone_contents_t

- only basic functionality with leaks

- sem checks missing, more extensive tests missing

- most checks can be done without allocating anything

- Lost my temper with the direct changes, rewritten DDNS to only
  changeset creation, application will be the same as usual.
- Will not work now

- not perfect, but without this there'd be no way to tell whether to
  free the owner or not.

updates.

- DDNS and changeset application will not work now, but it wouldn't work
  before either

- some cleanup and fixes as well, namely simplified RR creation during
  zone load (no malloc)

- RRSet copies removed, packet now stores rrsets directly (structure is
  copied, but copy is shallow)

- use rrset initialization macro instead, or create rrset explicitely

- kept RRSet copies somewhere (node has to create them now)
- replaced knot_rrset_t with knot_rrs_t elsewhere
- semantic checks simplified

- Will leak
- Untested

- new types: knot_rr_t and knot_rrs_t
- these can be used in the node structure instead of knot_rrset_t

No need to be picky about this, TSIG rcode is 16bit so we might
as well use 16bit number to keep things the same.

This function releases held RCU read lock, so it is possible that
the zone is free in the meantime. To prevent accessing the freed
zone, zone is released after all processing is done and the
pointer is invalidated to prevent further access.

The problem with previous code was that the packet stripped the TSIG
section when it parsed the first IXFR packet from wire. If the code
failed though, the wire was parsed again but this time the TSIG was
already stripped on the wire so the result was garbled data.
I've thrown in some minor cleanup as well.

This makes NSEC3 nodes disappear if it is preceded by a RR with the
same owner and any other record type.