fix(voice): full E2EE bidirectional audio pipeline working

- bot.py: track active callers per room; only stop session when last
  caller leaves (fixes premature cancellation when Playwright browser
  hangs up while real app is still in call)

- voice.py: pre-compute 8 HMAC-ratcheted keys from EC's base key so
  decryption works immediately without waiting ~30s for Matrix to
  deliver EC's key-rotation event (root cause of user→bot silence)

- voice.py: fix set_key() argument order (identity, key, index) at all
  call sites — was (identity, index, key) causing TypeError

- voice.py: add audio frame monitor (AUDIO_FLOW) and mute/unmute event
  handlers for diagnostics

- voice.py: update livekit-agents 1.4.2 event names: user_state_changed,
  user_input_transcribed, conversation_item_added

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

bot.py

@@ -266,6 +266,7 @@ class Bot:
         self.lkapi = None
         self.voice_sessions: dict[str, VoiceSession] = {}
         self.active_calls = set()  # rooms where we've sent call member event
+        self.active_callers: dict[str, set[str]] = {}  # room_id → set of caller user IDs
         self.rag = DocumentRAG(WILDFILES_BASE_URL, WILDFILES_ORG)
         self.memory = MemoryClient(MEMORY_SERVICE_URL)
         self.llm = AsyncOpenAI(base_url=LITELLM_URL, api_key=LITELLM_KEY) if LITELLM_URL else None
@@ -401,6 +402,7 @@ class Bot:
 
             logger.info("Call detected in %s from %s, joining...", room_id, event.sender)
             self.active_calls.add(room_id)
+            self.active_callers.setdefault(room_id, set()).add(event.sender)
 
             # Get the foci_preferred from the caller's event
             content = event.source["content"]
@@ -478,10 +480,20 @@ class Bot:
                     self.voice_sessions.pop(room_id, None)
 
         else:
-            # Empty content = someone left the call, check if anyone is still calling
+            # Empty content = someone left the call
             room_id = room.room_id
             if room_id in self.active_calls:
-                # Stop voice session
+                # Remove this caller from active set
+                callers = self.active_callers.get(room_id, set())
+                callers.discard(event.sender)
+                if callers:
+                    logger.info("Participant %s left %s but %d other(s) still in call — keeping session",
+                                event.sender, room_id, len(callers))
+                    return
+
+                # No callers left — stop voice session
+                logger.info("Last caller %s left %s — stopping session", event.sender, room_id)
+                self.active_callers.pop(room_id, None)
                 vs = self.voice_sessions.pop(room_id, None)
                 if vs:
                     try: