The pruning is going to be a little tricky with the db,
as of now first query that fills up cache to the brim clears it.

as glue (or any additionals) are NOT cached

- removed excess copies in query, cache time tracking
- simple insert/replace semantics
- zero-copy query if TTL recalculation is not needed
- doc

- cache is now fully optional, resolution works without cache as well
- proper mempool per each request, not mixed
- zone cut resolution doesn't need any per-context memory
- now it's possible to do qname minimization
- rplan keeps tabs on resolved queries
- added documentation etc.

- correct processing of NS records in the authoritative answers

- zonecut map is now truly a volatile cache to accelerate zone cut
retrieval, RDATA is stored in cache
- cache update is commited after each successfuly resolved query
- SBELT contains a static data as a fallback if the root server cannot
be found in the zone cut map, or the data is missing in the cache

this is a little hacked, I'd like to do list of nested
transactions later and short-lived read transactions.

TODO:
pruning, better removal, periodic cleanup, configuration

Slow start evaluates all NS in the preference list, using mean as
a sorting function. Invalidated NSs are pushed to the back of the
preference list.