refactor: composition root in lifespan, break core↔services cycle, shared data layer

The gateway was the composition root, device registry, inbound-call
policy, and call-operations service in one class, with core↔services
circular imports papered over by function-local imports, wiring done
by assigning private attributes, and MCP tools duplicating REST query
logic against their own sessions.

Composition:
- main.py's lifespan now builds every service and wires them by
  constructor/registration. gateway.from_config() is gone; core/ no
  longer imports services/ anywhere — the cycle is dead.
- Inbound-call policy moved to ReceptionistService.on_inbound_call
  (routing evaluation, reject/answer, screening dispatch); wired as
  the engine's on_incoming_call by the lifespan. Receptionist deps
  (tts/transcription/recording/routing) are constructor-injected —
  no more gateway._tts reach-through or importing hold_slayer's
  private _get_llm (now services.llm_client.get_llm, shared).
- Hold-slayer launch goes through a mode-handler registry
  (register_mode_handler); the gateway no longer knows the service's
  type. CallManager takes on_call_ended in its constructor.
- build_sip_engine() is a pure function taking explicit callbacks.
- api/routing.py uses the routing service from app.state via a
  proper dependency instead of gateway._routing.

Shared data layer:
- db.session_scope() is the one session convention (get_db wraps it).
- services/call_persistence.py gains the query/write functions and
  the single StoredCallFlow→CallFlow mapper; api/call_flows.py,
  api/call_history.py, and the six DB-touching MCP tools are thin
  wrappers over them — the two surfaces can't drift.
- legs_for_call() replaces the three private _call_legs scans
  (gateway transfer/hangup, REST dtmf, MCP dtmf).

7 new tests (mode-handler launch, on_call_ended hook, receptionist
inbound answer/reject, call-flow CRUD round-trip and history routes
against real SQLite through the shared layer). aiosqlite added to dev
deps for that.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

services/hold_slayer.py

@@ -23,35 +23,12 @@ from models.call import ActiveCall, AudioClassification, CallStatus, Classificat
 from models.call_flow import ActionType, CallFlow, CallFlowStep
 from models.events import EventType, GatewayEvent
 from services.audio_classifier import AudioClassifier
+from services.llm_client import get_llm
 from services.transcription import TranscriptionService
 from services.tts import TTSService
 
 logger = logging.getLogger(__name__)
 
-# LLM client is optional — imported at use time
-_llm_client = None
-
-
-def _get_llm():
-    """Lazy-load LLM client (optional dependency)."""
-    global _llm_client
-    if _llm_client is None:
-        try:
-            from config import get_settings
-            from services.llm_client import LLMClient
-
-            settings = get_settings()
-            _llm_client = LLMClient(
-                base_url=settings.llm.base_url,
-                model=settings.llm.model,
-                api_key=settings.llm.api_key.get_secret_value(),
-                timeout=settings.llm.timeout,
-            )
-        except Exception as e:
-            logger.debug(f"LLM client not available: {e}")
-            _llm_client = False  # Sentinel: don't retry
-    return _llm_client if _llm_client is not False else None
-
 
 class HoldSlayerService:
     """
@@ -228,7 +205,7 @@ class HoldSlayerService:
 
                 # Phase 2: LLM fallback if regex couldn't decide
                 if not decision and transcript:
-                    llm = _get_llm()
+                    llm = get_llm()
                     if llm:
                         try:
                             logger.info("🤖 Regex inconclusive, asking LLM...")