feat(voice): add on-device Miao Zhang Gui wake word

Add an optional local wake response prompt configuration. The prompt asset is intentionally not included yet.

components/platform/internal/multimodal/audio_pipeline.h

@@ -11,6 +11,10 @@ extern "C" {
 #endif
 
 typedef esp_err_t (*multimodal_audio_uplink_cb_t)(const uint8_t *data, size_t len, void *ctx);
+typedef void (*multimodal_audio_pcm_observer_t)(const int16_t *pcm,
+                                               size_t samples,
+                                               uint32_t sample_rate_hz,
+                                               void *ctx);
 
 typedef enum {
     MULTIMODAL_AUDIO_DOWNSTREAM_PLAYBACK_STARTED = 0,
@@ -33,6 +37,13 @@ esp_err_t multimodal_audio_start_uplink(multimodal_audio_uplink_cb_t cb, void *c
 esp_err_t multimodal_audio_stop_uplink(int timeout_ms);
 bool multimodal_audio_is_uplink_running(void);
 bool multimodal_audio_is_initialized(void);
+esp_err_t multimodal_audio_register_pcm_observer(
+        multimodal_audio_pcm_observer_t observer,
+        void *observer_ctx);
+esp_err_t multimodal_audio_unregister_pcm_observer(
+        multimodal_audio_pcm_observer_t observer,
+        void *observer_ctx,
+        int timeout_ms);
 
 esp_err_t multimodal_audio_handle_downstream_opus(const uint8_t *data, size_t len);
 esp_err_t multimodal_audio_begin_downstream_response(void);