This enables functionality when name is preallocated.

* When transaction security is required
  * Check it first and respond if fails
  * Always attempt to sign answers (if possible)
  * If TSIG required but not sent, return BADKEY and do not parse
    the prereq section - this differs from BIND impl., but it's more
    coherent.

Needs review though.
Occured in 29/12/12 nightly.

Conflicts:
	samples/knot.full.conf
	src/knot.conf.5
	src/knot/conf/conf.h
	src/knot/server/zones.c
	src/knot/server/zones.h
	src/knotc.8

This is really a workaround until correct automatic signing is
done and prevents crash in some scenarios.

fixes #2284
refs #2198

refs #2198
fixes #2285

refs #2176

This is probably a misconception. When transfer failed at some point,
the appropriate handling function would either send buffered response
and/or error message with specified rcode. But then the handling message
would free the response structure and caller of the function would attempt
to create and send ANOTHER error message based on the return value.
This would obviously fail hard, as the response structure is freed at this point.

refs #2176

Previously server terminated the connection, resulting in
nsupdate trying several times before giving up.
Should comply to RFC and also other servers behavior.

refs #2239

- Some variables were not initialized properly.
- Renamed rrset_compare() to rrset_match() as it does not return the
  values usual for compare functions.
- Fixed some potential leaks.
- Fixed removing redundant RRs from RRSets removed from node.

refs #2184

- Removing node from the zone tree if inserting to hash table fails.

refs #2184

- When RDATA have been merged to previously made copy of node's
  RRSet, the merged RDATA were not saved to the 'changes'
  structure and thus it did not get deallocated after failure.

refs #2184

refs #2184

refs #2184

- Would crash compilation with semantic checks on special zones.

Conflicts:
	src/knot/zone/semantic-check.c

Refs #2197

Conflicts:
	src/libknot/zone/zone-diff.c