refs #181

For when rescheduling itself with 0s interval just doesn't cut it.

refs #158

refs #180

When RRSIGs are carried in a separate packet, node and in_zone is
left initialized. This is legitimate.
I also suspect there are some cases where node is set to NULL but
in_zone is unchanged, but that's out of the scope of this patch.

- unified zone_sign and changeset_sign APIs - both use knot_zone_t * now as data input - fixes a problem with wrong zone->contents->zone link.
- Refuse NSEC3PARAM if owner is different from apex
- Do not resign removed RRSets
- Assertion is_valid_nsec3_node turned to check - there are several cases where this will fail
- log message fixed
- Do a whole zone resing after NSEC3PARAM change - not needed for all cases, but NSEC3PARAM addition might make deactivate some DNSKEYs (RSASHA1 algo), and then the zone would be invalid

Meanwhile, fixed the most obvious race conditions with reload.

refs #125

Conflicts:
	src/knot/server/zones.c

refs #125

refs #125

Fix unchecked output value of snprintf() detected by Coverity.

CID #1140397

refs #158