button_relay_toggle_v2: bound retained-command guard to a time window, fix active_low re-drive

[KrX3D]

Summary

Follow-up review of button_relay_toggle_v2 (after #7/#25), looking for anything left. Found and fixed two real bugs, plus a doc typo:

• The stale-retained-command guard could swallow genuine live commands. _bootOverridePending[i] gets armed on every settings save and MQTT reconnect via setupRelaySubscriptions() (for every relay with a valid pin, not just non-Default boot-state ones), and stays armed until the next /set message arrives — however much later that is. If a real, live Home Assistant command happened to be the first one received after a reconnect that occurred minutes earlier, it was misclassified as "the stale retained redelivery" and silently overridden back to the relay's current state, dropping the user's command. Replaced the boolean with a short (3s) time-bounded deadline (_overrideDeadline[i]) — retained messages are redelivered by the broker essentially immediately on subscribe, so this window still catches that case without swallowing an unrelated later command.
• Toggling relay_active_low alone silently flipped the reported state. In readFromConfig()'s "same pin" branch, when only relay_active_low changed (pin/wiring untouched), the code called pinMode(OUTPUT) but never re-drove the pin. Since the reported logical on/off state is derived as activeLow ? !digitalRead(pin) : digitalRead(pin), flipping the setting flipped what HA/MQTT reported even though the relay's actual electrical state never changed. Now captures the old active_low value before the config read and, if it changed, re-drives the pin so the logical state stays the same as before the save.
• Fixed a corrupted arrow character (? instead of ->) in the readme's wiring diagram.

I looked at a third finding from the review — plumbing the MQTT retain flag through to onMqttMessage() instead of approximating it by timing — but that requires changing the Usermod::onMqttMessage() interface used by every usermod (wled00/mqtt.cpp, wled00/fcn_declare.h), which is out of scope for a usermod-only fix. Noting it here as a possible follow-up for core.

Test plan

• Static review only in this session (no PlatformIO toolchain available to build/flash here).
• Recommend testing on real hardware: toggle relay_active_low alone via settings and confirm HA doesn't see a spurious state flip; and confirm a live HA command right after a settings save/reconnect is no longer overridden.

🤖 Generated with Claude Code

https://claude.ai/code/session_01J6bJtispewBKR2dXBcNWpw

---

Generated by Claude Code

[KrX3D]

The wled_build / Test cdata.js check failed on the first run, but it's not related to this PR's diff (which only touches usermods/button_relay_toggle_v2/button_relay_toggle_v2.cpp and its readme.md). The failing assertion is in tools/cdata-test.js ("script was executed with -f or --force" expecting wled00/html_ui.h to have been touched by a forced rebuild) — a filesystem-mtime-based test with no connection to this change, and the same commit history's most recent direct push to button_relay_toggle_UM passed this exact check. Re-running the failed job once to confirm it's a flake.

---

Generated by Claude Code

[KrX3D]

Re-run passed — confirms the earlier Test cdata.js failure was the flaky mtime-based subtest, unrelated to this change. CI is green.

---

Generated by Claude Code