[BUG] minimal mDNS responses / announcements for unique records (A, AAAA, SRV, TXT) don't set the cache-flush bit

[ksperling-apple]

Reproduction steps

1. Run e.g. all-clusters-app on ESP32/M5Stack (with CONFIG_USE_MINIMAL_MDNS=1)
2. Observe mDNS packets on the network

E.g.

Multicast Domain Name System (response)
    Transaction ID: 0x0000
    Flags: 0x8400 Standard query response, No error
    Questions: 0
    Answer RRs: 4
    Authority RRs: 0
    Additional RRs: 3
    Answers
        _services._dns-sd._udp.local: type PTR, class IN, _matter._tcp.local
        _services._dns-sd._udp.local: type PTR, class IN, _I49897676B257FAF4._sub._matter._tcp.local
        _matter._tcp.local: type PTR, class IN, 49897676B257FAF4-000000000000022B._matter._tcp.local
        _I49897676B257FAF4._sub._matter._tcp.local: type PTR, class IN, 49897676B257FAF4-000000000000022B._matter._tcp.local
    Additional records
        49897676B257FAF4-000000000000022B._matter._tcp.local: type SRV, class IN, priority 0, weight 0, port 5540, target 782184957AAC.local
        49897676B257FAF4-000000000000022B._matter._tcp.local: type TXT, class IN
        782184957AAC.local: type AAAA, class IN, addr fe80::7a21:84ff:fe95:7aac
    [Unsolicited: True]

In this example, the SRV, TXT, and AAAA records should be considered part of "unique" RR set and hence have the cache-flush bit set. Note that as per RFC 6762 advertising unique RR sets also requires probing which appears to not currently be implemented in the minimal mDNS implementation.

One practical scenario where we've seen this have an impact is when retrying a commissioning flow that brings the device onto the network. It's often the case that the IPv6 LL address becomes available some seconds before a routable IPv6 address with a prefix announced by the router. In this case the device will send an announcement (without cache-flush) containing it's IPv6 LL address, but the commissioner may still have a previous AAAA record with a routable address cached, and prefer it to the valid LL address. The commissioner then ends up retransmitting the CASE Sigma1 message until the ULA actually becomes available or the CASE establishment times out.

### Bug prevalence

100% reproducible

### GitHub hash of the SDK that was being used

0b8ecc84

### Platform

esp32, other

### Platform Version(s)

_No response_

### Anything else?

_No response_