perf(models): start from the cached catalog and refresh in the background

Every host start awaited a full catalog request before the provider
registered, costing one HTTPS round-trip (up to the discovery timeout on a
hanging connection) even when a cache written seconds earlier was on disk.

Register the cached catalog immediately and run the live refresh in the
background; the live result re-registers the provider when it arrives. A
first start without a cache still awaits the live catalog. The background
refresh is aborted on session_shutdown so print mode does not wait for it.

Closes #63

(cherry picked from commit 142191f9420fe90460cdc3cbae70e26d0c000860)

index.ts

@@ -29,6 +29,7 @@ import {
   DEFAULT_PROVIDER_API_BASE,
   getModelsTimeoutMs,
   inputModalitiesForModel,
+  loadCachedCommandCodeModels,
   loadCommandCodeModels,
   MODEL_EFFORTS,
   thinkingMetadataForModel,
@@ -159,15 +160,21 @@ export default async function (pi: ExtensionAPI) {
   const runtime = createCommandCodeRuntime<ProviderConfig, ExtensionCommandContext>(pi, {
     endpoint: modelsUrl,
     cachePath: modelsCachePath,
-    loadModels: () =>
+    loadModels: (signal) =>
       loadCommandCodeModels({
         url: modelsUrl,
         cachePath: modelsCachePath,
         timeoutMs: modelsTimeoutMs,
+        signal,
       }),
+    loadCachedModels: () => loadCachedCommandCodeModels(modelsCachePath),
     createProviderConfig: (models) => createProviderConfig(models, apiBase, transport.stream),
     getTransport: transport.getTransport,
   })
 
+  pi.on("session_shutdown", () => {
+    runtime.dispose()
+  })
+
   await runtime.initialize()
 }