- Deleted obsolete files
- Added new b64 decode/encode implementation (same author as out current b32 impl.)
- Makefile/Knot.files changes.

- When not-in-zone dname was present in RDATA for more than one time
  and was covered by wildcard at the same time, wrong node was freed.

Now everywhere the server tries to respond with the whole Question
section. When it is not possible, it at least copies the ID and
flags.

refs #1935 @30m

Not only the name but also the algorithm must be checked.

refs #1933 @10m

An ugly bug that could have manifested in many other ways.

refs #1933 @30m

Conflicts:
	src/libknot/nameserver/name-server.c

Refs #1419 @3h

refs #1909

refs #1688

refs #1909

- When IXFR is rejected and AXFR is tried, wrong incoming packet
  could have crashed the server. Fixed this by cleaning the
  ns_xfr_t structure before the fallback.

refs #1912

refs #1688

Old: [AI]XFR of 'ZONE/(IN|OUT)' ...
New: (Incoming|Outgoing) [AI]XFR of 'ZONE' ...

refs #1912

refs #1912

- Fixed check in RDATA adjusting.
- Fixed searching for wildcard covering RDATA dname.
- Fixed zone_contents_find_dname().

TODO: Leaks memory from some nodes created during load, but only
      when there are some RDATA dnames covered by a wildcard in
      the zone.

refs #1419

- In both cases check also if the NSEC3 tree is not empty. That
  may cause problems.

refs #1922, #1933, #1597

refs #1922, #1923

refs #1912

refs #1912

Conflicts:
	src/common/dynamic-array.c
	src/knot/server/zones.c

refs #941

Wondering whether it would do everything as it should.

refs #941

Conflicts:
	src/libknot/nameserver/name-server.c

TODO: Use the new functions in all possible cases, there are more
      possible, mostly where ns_error_response() is now used.

refs #1918 @1h