maradns: parse cname and unknown raw records

[aberoham]

Two test cases were disabled with it.skip. Both hid real parser gaps.

cname was missing from the set of type tokens the parser recognises, so a CNAME line was read as an A record. dname and uri had the same problem. The dispatch switch already had cases for all three. None was ever reached, so all three threw:

before:  sub.example.net. DNAME target.example.net. ~   -> throws (read as A, fails RFC 1035)
after:   -> { owner, ttl, type: DNAME, target: 'target.example.net.' }

URI needed one more thing. csv2 quotes the target, and those quotes are what stop expandPercent from eating a percent-encoded URI, so they can only come off afterwards. The target now matches what fromTinydns produces:

_http._tcp.example.com. URI 10 1 'https://example.com/a%20b?x=1' ~
-> target: 'https://example.com/a%20b?x=1'

A RAW line with an unrecognised type id returned false and the record was discarded. It now returns an RR.UNKNOWN per RFC 3597. This changes output. A zone containing such a record now yields one more record than before. The fixture count moves from 40 to 41.

A malformed type id still returns false. RR.UNKNOWN validates the id and throws invalid DNS type: TYPENaN, so without that check a line that used to be skipped would end the parse instead.

RR.UNKNOWN arrived in dns-resource-record 1.9.0. That is already the declared floor.

231 tests, 231 pass, 0 failed, 0 skipped