pure hashtable is a node type where all keys share a common letter,
so it is consumed on parent node -> pure trie node transition.
So when tracing back, this letter must be returned.

Also updated unit tests to cover these sort of situations.

- reorganized
- some code blocks are split
- SIGALRM handler for main thread

- previously static function 'put_rr' in internet.c has been made public, it handles additions of RRSIGs as well now, by adding the to a list that is traversed after each section has been solved.

- Removed rrset->rdata_indices, RDATA lengths are now stored inline together with RR's TTL
- Added function to get RR's TTL - TTL of the whole RRSet can still be obtained, but it will default to the first RR's TTL or 0
- Renamed few RRSet functions, removed a few too - the file still needs a cleanup and function reordering
- Special handling for RRSIGs during AXFR-out no longer needed (yay)

throughout

- it is as good as the sockaddr_t previously, but standard
- subnets are only needed for ACLs
- common way to print addresses (even AF_UNIX)

- refactored the zone-loader a bit, as reading from zone file
  creates only contents and the zone is created separately
- refactored zone contents creating function, as it is always used in a
  way where an apex node is created and inserted into contents
- crash occured because created zone bound configuration and if parsing
  failed, the configuration was freed with the zone and also with the
  caller

* events now carry pointer to their scheduler
* no event types, events are simple callbacks
* removed unused or deprecated stuff
* bugfix: zone never expired since expire was initialized on create
* bugfix: race condition on rescheduling events
  - there is always some time between cancel and schedule
  - if two threads run run past the cancel and then race for schedule,
    they may rarely schedule an event two times
  - this is hard to deal with unless we place a lock on the zone
    operations, therefore each schedule checks heap if the event is not
    already scheduled and if it is, it replaces the old instance

* removed redundant attributes
* removed unused parameter occurences
* moved related functions to appropriate files
* evsched has now context (server)
* zone does not point to server
* zone event planning is still in the zone loading and requires server #209

also includes removal of knot_ prefixes from private API

close #205

issue #199

Simpler knot_pkt_reserve() API should be used instead.

issue #1

Resizing of thread units on the run is not used nor endorsed,
as it's hard to guess when the specific threads terminate or not.
(They are allowed to run to the nearest cancellation point or finish.)