1.) have write_rrsets() return all RRsets value (idempotence)
Previously, only new or changed RRsets that were written to pdns had
been returned. Now, all RRsets that exist at the end of the operation
are returned (i.e. only the deleted ones are missing).

2.) dummy TTL for RRsets up for deletion

3.) fix deadlock by rephrasing WHERE clause for RR deletion

4.) refactor: simplify and clarify Domain.write_rrsets()

This commit also slightly shortens certain cache times for production, from
default to 20s.

This commits refactors chakram.resolve() and introduces chakram.resolveStr(),
both return promises that will eventually be resolved into response data from
nslord.

It also introduces dns, a dummy property to increase readability of
expectations and ttl(exp), a function to check if a RR set has the correct ttl.

1.) remove unnecessary flags from DNS resolve code
The following flags (i.e. all) were removed:
- Recursion Desired: we don't want nslord to recurse
- Authoritative Answer: reserved for responses, see RFC 1035 Sec. 4.1.1

2.) close socket after tests so that mocha will exit

3.) allow NXDOMAIN DNS response code
This is a valid response code when looking for a non-existing RRset
on a subdomain that has no records at all.

4.) add TXT record serialization to DNS toolset

5.) augment TTL test method

Previously, IBAN and BIC were normalized and stripped from
whitespace only for creating the direct debit attachment file.
If the user entered spaces like '   DE123455....', those were
stored in the database. Along with the 6-character cutoff, this
was not desirable.

Space stripping is now done in the Donation serializer so that
it applies to both what's stored and to what's put in the direct
debit file.

Also, e2e tests were adapted to verify the presence of the masked
IBAN.

OPT is a reserved record type with special meaning and is not supposed
to exist in a zone. Currently, if an OPT RRset is created, the zone may
break.

Until PowerDNS ensures that RRsets of type OPT cannot created, let's
make sure on this level.

PowerDNS issue: PowerDNS/pdns#6441

pdns would convert types like TYPE99 to the named ones (here: SPF),
but our API cannot do so without maintaining a mapping table that
depends on the pdns version. Currently, we would store two distinct
RRsets in our API database, corresponding to only one RRset in pdns.
Worse, if one RRset was deleted (along with the one on pdns), the
other would remain as an orphan.

The only usecase for this would be new types quickly gaining
popularity although pdns does not know them yet. In this case, we
can address the issue again.