This reverts commit 243d4ac1.

This reverts commit 756debd0.

Added some missing read locks.
TODO: may deadlock somewhere, some functions still find zone from zonedb
and may corrupt new reloaded zone instead of zone the transfer was started

refs #1976

refs #1976

- Actual RDATA of zone were beeing changed due to improper API usage.

Refs #1965

- Missing NSEC/NSEC3 is no longer considered an error.

refs #1970

refs #1950

- If a node contains a CNAME RRSet or just a RRSIG covering a
  CNAME, the server will not try to resolve the CNAE chain.

refs #1971

This option allows binding to the same port using separate
IPv6 and IPv4 socket.

refs #1877

refs #1877

- Freeing RRSIGs after RRSet deep copy (there's no switch that would mean no RRSIGs copyiing)
- Added assert to zone-dump that checks for such state

Refs #1752

refs #1968

refs #1969

- Added to zones section (configuration reference), might be moved later

Refs #1877 @2h

- Was causing compilation error with latest OpenSSL (v1.0.1C)
- Made rdata conversion functions static.

- Disabled where I was able to tell it's OK.
- Enabled in zone-diff, as I'm not sure it would work without it.

Should mitigate problems with IXFRs.

Conflicts:
	src/libknot/updates/xfr-in.c

- Wrong flag was used.
- Added some detailed debug messages (may be removed later).

refs #1952 @3h

- If the first RRSet was empty and the second one contained only
  one RDATA, this RDATA would get deallocated.

- Storing each RDATA separately during changeset application. The
  chains may get corrupted due to other changes and thus may not be
  deallocated properly. This caused leaks in some cases when the
  IXFR failed.
  This includes counting how many space we must reserve by counting
  number of RDATA to be stored.
- Function for getting RDATA count from the RDATA chain itself

- Returning new node or NULL (if empty) in dname_get_node().
- Modified dname_update_node() to consider empty nodes.
- Freeing empty nodes after everything is finished.
- Marking empty nodes by a new flag.