fix(mdns): Fix mdns browse SYNC vs. END race

[david-cermak]

1)

followup fix on the recently merged #1123
(adds a similar quick fix as in 9057f2d)

2)

Update CI tests/IDF version matrix

3) 1.12.0

Features

• add hostname change callback (c2cd90b0)
• Make mDNS browse feature optional (2e0c2540)

Bug Fixes

• Update supported/tested IDF version matrix (e8427217)
• Fix mdns browse SYNC vs. END race (1b83e978)
• Only depend on esp_wifi when Wi-Fi is enabled (47724b7d, [esp_wifi seems to be always included even when only using ethernet (IDFGH-15610) #835](https://github.com/espress
  if/esp-protocols/issues/835))
• Update host test to check goodbye and browse dupl (ab03980d)
• Fix mdns_browse_new() return dangling pointer on duplicated browsers (e063449c)
• Create static task only when SPIRAM-task-alloc is ON (6f31f27f)
• validate subtype label position (a09b24fe)
• ignore subtype SRV and TXT questions (7910501d)
• distinguish subtype queries from instances (b9103169)
• Don't free a browse result already freed by an earlier sync batch (9057f2d9)

---

Note

Medium Risk
The browse sync guard touches core mDNS action-queue concurrency; incorrect checks could drop valid notifications, but the change is narrowly scoped and aligns with an existing result-pointer pattern.

Overview
Fixes a use-after-free race in mDNS browse handling when ACTION_BROWSE_END detaches and frees a browse while a queued ACTION_BROWSE_SYNC still holds a borrowed pointer. browse_sync() now verifies the browse is still on s_browse and in BROWSE_RUNNING before delivering notifier callbacks (mirroring the existing result_is_cached guard for result nodes).

CI updates expand ESP-IDF coverage: target build matrix moves to release-v5.5 / v6.0 / v6.1 with adjusted expected Kconfig warnings for v6+, and host tests run against a matrix of IDF versions instead of a single release-v5.3 image.

Host test app changes improve Linux console pytest reliability: exit calls exit(0) on Linux to avoid hung REPL threads, and IDF ≥ 6.1 uses esp_console_new_repl_stdio instead of UART REPL on the Linux target.

^{Reviewed by Cursor Bugbot for commit 41bffa0. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit c0282cf. Configure here.}

[euripedesrocha]

Only a minor comment on host test matrix.

[euripedesrocha]

Why keep a matrix in the host tests?

[david-cermak]

Good point, not really necessary, but I was thinking in CI resources -- the host tests are cheap/quick and mdns code has some IDF_VERSION ifdefs, so it's yes another (fast) feedback if something broke.