docs: changelog for 2.16-beta.5 (HW4 nag escalation-edge fix + honest AP-First banner)

Author: hypery11

changelog.md

@@ -1,3 +1,9 @@
+## 2.16-beta.5 — HW4 nag killer: escalation-edge grip pulse + honest AP-First banner (testing build)
+
+- **HW4 Juniper nag killer fix (#100)** — beta.4 suppressed the first hands-on nag wave but the second still escalated to yellow on HW4 Juniper trims. Root cause: on those trims EPAS handsOnLevel (`0x370` byte4) is frozen at 0 the whole hands-off window (measured constant by @jewelrylin / @DrStrangeglovebox), so the strong grip pulse — which edge-detected only handsOnLevel — fired once and never re-armed. The signal that *does* move is `das_hands_on_state` stepping 0→2→3. The grip pulse now re-arms on a DAS escalation rising edge too, tracking das every frame so a satisfied (0) wave resets the baseline. It only changes *when* the pulse fires, not the gate or the echo, so HW3 and steady-das HW4 cars are unchanged. The ESP32 nag path had no on-demand pulse at all — added it with the same DAS-edge re-arm so both builds match. Host-tested; needs on-car confirmation across two consecutive nag waves.
+- **Honest AP-First banner on the ESP32 dashboard (#52/#108)** — the 14.x warning told users to "enable AP-First in settings", but AP-First is **Flipper-only** — there is no such control on the ESP32. Reworded to say so and give the manual workaround (engage AP from the stalk first). Porting the AP-First gate to the ESP32 is tracked separately.
+- Thanks: @jewelrylin and @DrStrangeglovebox (the dual-unit HW4 nag captures that diagnosed this), @kristopf007, @dunckencn.
+
 ## 2.16-beta.4 — Continuous AP (HW3), HW4 nag fallback, 0x229 cracked + gated (testing build)
 
 - **Continuous AP (HW3/Legacy, #107, @vrs11)** — enhAuto-style auto-re-engage: while Autopilot is active it watches the disengage conditions, and if AP drops with a turn signal on it waits for low steering torque + signal off + AP ready, then re-engages (max 3 attempts). Brake pedal or a full right-stalk-up is the kill switch, re-checked in every state. **Opt-in (off by default), HW3/Legacy only** — the HW4 engage sequence isn't identified yet so the HW4 path is stubbed inert. The brake interlock fails closed: it won't engage until a `0x145` (ESP_status) frame has been seen.