fix(inference): narration model_id accepts bare provider id from TOML

The router matched only the adapter-prefixed registry id
(llama-cpp:narrator) while the operator's [router.narration] model_id
naturally repeats the bare [[providers]] id (narrator), so the match
always failed and narration silently fell back to the slow main model.
Match either form.

core/inference/src/main/kotlin/com/correx/core/inference/DefaultInferenceRouter.kt

@@ -69,7 +69,11 @@ class DefaultInferenceRouter(
 
         val all = registry.listAll()
         val healthy = all.filter { cachedHealth(it) !is ProviderHealth.Unavailable }
-        val matched = healthy.firstOrNull { it.id.value == modelId }
+        // Registry ids are adapter-prefixed ("llama-cpp:narrator") while the operator's TOML
+        // declares the bare id ("narrator") — accept either form.
+        val matched = healthy.firstOrNull {
+            it.id.value == modelId || it.id.value.substringAfter(':') == modelId
+        }
         if (matched == null) {
             log.warn(
                 "narration model_id '{}' matched no healthy provider — falling back to capability routing",