All data in server are kept in canonical format after recent
changes.The flag is no longer needed.

Existing code didn't recognize case sensitive RR types (LP),
which caused creation of invalid signatures.

RDATA lowercase

Some more tests would be handy, will add next week.

See merge request !288

Code cleanup

See merge request !290

EDNS fixes

- Fixed interpretation of Extended RCODE (was not joining the normal and extended RCODE fields).
- Modified output of Extended RCODE in utilities.
- If BADVERS is returned, return minimal error response. (No Answer)

See merge request !289

- Test changing case in different types and even adding a duplicate
  record that differs only in lettercase in RDATA.
- Also, test changing case in NSEC's next domain name. Should result
  in resign, but that doesn't happen right now.

- Semantic check: missing conversion of next NSEC to lowercase
  when comparing with apex.
- Changed condition when SERVFAIL is returned (no combination
  with any possible Ext RCODE).

Buffer boundaries are checked in knot_dname_wire_check(). The tested
function expects a valid dname. As a result, a memory outside the static
buffer is accessed.

_BSD_SOURCE is deprecated since glibc 2.20

Robert Edmonds authored 10 years ago and Jan Včelák committed 10 years ago

protobuf-c 1.0.0 added a new 'allocator' field to ProtobufCBufferSimple
that controls memory allocation, which must be NULL in order to request
the default system allocator. Allocating ProtobufCBufferSimple objects
on the stack without zeroing the entire object can result in
protobuf-c's memory allocation functions dereferencing a garbage
pointer.

Signed-off-by: Jan Vcelak <jan.vcelak@nic.cz>

Conflicts:
	Knot.files

Only RFC 4034 types are converted to lowercase.
Unified the conversion to one RRSet function used in both zone
parsing and packet parsing.