the overlay may be used for cached data now, and it's not okay to reparse
it if not needed or wanted

Enable SILENT_RULES in SPHINX build

This patch will make builds much less verbose and more verbose at the same time according to SILENT_RULE settings (--enable/disable-silent-rules or make V=0/V=1), see:

```
$ make clean all html
rm -rf .libs _libs
rm -rf ./_build
rm -f man/*.1 man/*.5 man/*.8
rm -f *.lo
  GEN      man/knot.conf.5
  GEN      man/knotc.8
  GEN      man/knotd.8
  GEN      man/kdig.1
  GEN      man/khost.1
  GEN      man/knsupdate.1
  GEN      man/knot1to2.1
  GEN      man/knsec3hash.1
  GEN      man/keymgr.8
  SPHINX   html-local
```

See merge request !374

Reimplementing the daemon() call for OSX targets and to support older
setups. Daemonization today is deprecated altogether since the init
system (systemd,rc,upstart,...) should take care of process
daemonization themselves. This just mutes the warning.

Proper integer encoding in signature and public key conversion operations.

Closes #357 #358

cherry pick from the libknot cleanup wip, this is going to be used
much more in the future for storing layer module owners both in
auth and resolver + for modules with ffi

Packet's wire memory initialization

Valgrind reported massive ammounts of conditional jumps depending on uninitialized memory in the security/protos test. This fixes it and as a bonus removes excess memset and reworks packet initialization a little bit.

See merge request !367

packet: bugfix: revert packet's wire initialization and fix it properly; knot_rrset_rr_from_wire read OOB.

Fixes #359

The previous knot_map_errno was vulnerable to stack buffer undeflow
if the input list was not terminated with a zero.

Fix building of documentation out of the source tree.

Don't release tar.gz, just tar.xz

I think it's quite safe to drop tar.gz these days and just distribute tar.xz.  What do you all think? @jvcelak @dsalzman @dtaborsky ?

See merge request !368