fix: commit pending host drift, enable emotive ack rotation

docker-compose.override.yml: VAD tuning (RMS threshold, min speech, trailing
silence), no-speech reprompt, immediate_short ack mode, media idle timeout --
all previously live on prod but uncommitted. AI_VOICE_V2_EMOTIVE_ACK_ENABLED
flipped 0 -> 1: this flag was silently disabling filler-phrase rotation
regardless of the app-level 36c2acb fix.

docker-compose.minimal.yml: removed the voice-ai (Magzhan legacy) service
and its dependency, matching what was actually running.

docker-compose.yml: added common_network to api-gateway.

docker-compose.override.yml

@@ -8,13 +8,15 @@ services:
     - AI_VOICE_V2_DUPLEX_ENABLED=1
     - AI_VOICE_V2_STREAMING_ASR_TIMEOUT_SECONDS=15
     - AI_VOICE_V2_STREAMING_ASR_SUPPORTED_LANGUAGES=ru,kk
-    - AI_VOICE_VAD_RMS_THRESHOLD=800
-    - AI_VOICE_VAD_MIN_SPEECH_MS=300
-    - AI_VOICE_VAD_TRAILING_SILENCE_MS=500
+    - AI_VOICE_VAD_RMS_THRESHOLD=300
+    - AI_VOICE_VAD_MIN_SPEECH_MS=200
+    - AI_VOICE_VAD_TRAILING_SILENCE_MS=800
+    - AI_VOICE_NO_SPEECH_REPROMPT_ENABLED=1
+    - AI_VOICE_V2_ACK_MODE=immediate_short
+    - AI_VOICE_V2_EMOTIVE_ACK_ENABLED=1
+    - AI_VOICE_MEDIA_IDLE_TIMEOUT_SECONDS=20
     ports:
     - 127.0.0.1:9019:9019
-    volumes:
-    - /tmp/asr_modified.py:/app/services/ai_voice_runtime_service/providers/asr.py:ro
   asterisk-bridge-service:
     environment:
     - ASTERISK_AMI_HOST=172.19.0.1