feat(router,inference,config): pin narration to a configured model_id

[router.narration] model_id selects which provider handles narration turns
instead of generic capability routing — lets a small/fast model own narration
while the main model keeps CHAT/STEERING.

- NarrationSettings.modelId parsed from TOML, threaded via RouterConfig
  .narrationModelId into DefaultRouterFacade.narrate (3-arg route)
- DefaultInferenceRouter: exact-id match against healthy providers, with a
  post-select health check; any miss logs WARN and falls back to capability
  routing (never fails the narration turn)
- onUserInput unchanged — only narrate uses the pinned model (tested)

core/router/src/main/kotlin/com/correx/core/router/RouterFacade.kt

@@ -175,7 +175,7 @@ class DefaultRouterFacade(
 
         val contextPack = routerContextBuilder.buildNarrationContext(state, trigger, config.tokenBudget)
 
-        val provider = inferenceRouter.route(effectiveStageId, setOf(ModelCapability.General))
+        val provider = inferenceRouter.route(effectiveStageId, setOf(ModelCapability.General), config.narrationModelId)
         val inferenceRequest = InferenceRequest(
             requestId = InferenceRequestId(UUID.randomUUID().toString()),
             sessionId = sessionId,

core/router/src/main/kotlin/com/correx/core/router/model/RouterConfig.kt

@@ -24,4 +24,5 @@ data class RouterConfig(
         topP = 0.9,
         maxTokens = 4096,
     ),
+    val narrationModelId: String? = null,
 )