Export TXTEntry so callers can set TXT records on advertised services

[mcuelenaere]

Summary

The DNS-SD server support (#253) lets callers advertise a service with WithService, but ServiceInstance.Text used the unexported txtKeyValue element type — so external callers had no exported type, constructor, or setter to populate TXT records. The server already serialized Text and Browse() already decoded into it; the only missing piece was the public API to construct entries.

This renames txtKeyValue → the exported TXTEntry and adds two ergonomic constructors that hide the nil-vs-empty Value footgun.

Before

// Impossible from outside the package — txtKeyValue is unexported.
mdns.WithService(mdns.ServiceInstance{
    Instance: "My Web Server",
    Service:  "_http._tcp",
    Port:     8080,
    // Text: ??? no exported element type
})

After

mdns.WithService(mdns.ServiceInstance{
    Instance: "My Web Server",
    Service:  "_http._tcp",
    Port:     8080,
    Text: []mdns.TXTEntry{
        mdns.NewTXTEntry("version", "1.2.3"), // "version=1.2.3"
        mdns.NewTXTEntry("setup", ""),         // "setup="
        mdns.NewTXTFlag("secure"),             // "secure" (boolean attr)
    },
})

Backward compatibility

Fully backward-compatible: txtKeyValue was never exported, so no external caller could reference it. Existing validation (non-empty key, 255-byte per-string limit, case-insensitive dedup keeping first) is unchanged.

Tests

• TestNewTXTEntry / TestNewTXTFlag — constructor semantics, incl. explicit empty value (key=) vs boolean flag (nil → bare key)
• TestNewTXTConstructors_Encode — entries → wire strings → decode round-trip
• TestCreateServiceAnswerTXTWithEntries — server emits the configured TXT entries in its TXT answer
• e2e testReverse now advertises a TXT entry + flag against avahi-browse

Refs #253

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.62%. Comparing base (2575a81) to head (a54bccf).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #277      +/-   ##
==========================================
+ Coverage   80.57%   80.62%   +0.04%     
==========================================
  Files           7        7              
  Lines        1596     1600       +4     
==========================================
+ Hits         1286     1290       +4     
  Misses        201      201              
  Partials      109      109              

Flag	Coverage Δ
go	80.62% <100.00%> (+0.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[backkem]

nits:

1. NewTXTEntry/NewTXTFlag don't validate keys (empty, contains =, non-ASCII). Validation only fires at encode time and even then only checks empty key + length. Pre-existing gap, but constructors are the natural place to add it.
2. NewTXTEntry only accepts string values. RFC 6763 ┬º6.5 allows opaque binary. Callers can still use TXTEntry{Key: "k", Value: bytes} directly, but worth a godoc note.
3. testReverse adds TXT entries but never asserts on them in the avahi-browse response. The e2e change is cosmetic as-is.

I think only the 1st one is worth considering changing before we merge because it would require changing the signature of NewTXTEntry/NewTXTFlag to add in a follow-up. What do others think?

[jwetzell]

Something else worth noting about text entries (that is kind of a pre-existing problem) is that keys are not case sensitive. Not sure how this should play out in practice since right now this is just a slice so we don't have any way to force this behavior until encode time.