A lot of syntax fixes (sorry, was not compileable before).

refs #88

Tests implemented so far:
- create empty
- create non-empty
- set_item() for setting RDATA items one-by-one

TODO:
= set_items()

refs #88, #31

Tests were not freeing memory.
Replaced forgotten printf by debug routine in dname.c.

refs #88, #31

- creating RDATA structure (empty, possibly with reserving space)
- setting and getting particular items
- setting all items (only if empty)
- deallocating

refs #5, #88

Tests of all API functions implemented, except for
  dnslib_dname_free() - no way to test that yet (consider using
  storage allocation hooks:
  http://www.cs.utah.edu/dept/old/texinfo/glibc-manual-0.02/library_3.html#SEC30
  )

refs #5, #31

dname and node were referencing each other. Used
  forward-declaration of node in dname.
Added API getter functions (with implementation):
  - dnslib_dname_name()
  - dnslib_dname_size()
  - dnslib_dname_node()
Fixed dname implementation:
  - str_to_wire() was replacing chars by label lengths
  - new_from_str() bad assert
  - new_from_wire() was not allocating space for the name
  - to_str() was de facto copying the wire format

Added some debug output to dname

refs #5, #88

refs #5, #88

Fixed creating dname struct from wire format (was not copying).
refs #5, #88

- creating dname (empty, from string, from wire format)
- deleting dname
- converting dname to string

Minor change in node - added linked list pointer to next node.

refs #5, #88

refs #5, #88

refs #5

* Test now detects invalid number of parameters.

Tests for server unit.
TAP protocol output.

- No epoll and master thread
- Workers are racing on recvfrom()
- Workers are natively blocked by recvfrom()
- Zero synchronisation overhead
- Presuming recfrom() is thread-safe