feat: correx-managed model lifecycle slice 5 — Go TUI model swap + telemetry

Go TUI decodes model.changed / model.list / resource.status and renders a
status-bar VRAM/GPU%/RAM gauge plus a models overlay (m key / palette 'models'):
lists configured models with the resident one marked, enter swaps (SwapModel),
c clears the pin (ClearModelPin). All three new server messages are
non-event-bearing control/gauge frames, applied immediately like
provider.status_changed.

Server addition required for the picker: ServerMessage.ModelList (model.list,
NonEventMessage) sent once in the initial snapshot from
ManagedInferenceRouter.availableModelIds()/currentModelId() — the TUI otherwise
cannot enumerate swap targets.

Completes the 5-slice model-lifecycle feature. ./gradlew check green; go build/vet clean.

Plan: docs/plans/2026-05-31-model-lifecycle-management.md (slice 5 of 5).

apps/server/src/main/kotlin/com/correx/apps/server/protocol/ServerMessage.kt

@@ -283,6 +283,19 @@ sealed interface ServerMessage {
         override val sessionSequence: Long? = null,
     ) : ServerMessage, NonEventMessage
 
+    /**
+     * The configured local models and which one is currently resident. Sent once in the initial
+     * snapshot when correx manages the model process, so clients can offer a swap picker.
+     */
+    @Serializable
+    @SerialName("model.list")
+    data class ModelList(
+        val models: List<String>,
+        val current: String?,
+        override val sequence: Long? = null,
+        override val sessionSequence: Long? = null,
+    ) : ServerMessage, NonEventMessage
+
     /**
      * Live resource gauge pushed periodically on the global stream (not event-derived). All fields
      * are nullable: GPU fields are null on a non-NVIDIA host, `processRssMb` is null when no model

apps/server/src/main/kotlin/com/correx/apps/server/ws/GlobalStreamHandler.kt

@@ -149,6 +149,16 @@ class GlobalStreamHandler(private val module: ServerModule) {
             ServerMessage.WorkflowList(workflows = workflows),
         )))
 
+        // Configured models + current, so the client can offer a swap picker (managed path only).
+        module.modelSwapper?.let { swapper ->
+            session.send(Frame.Text(ProtocolSerializer.encodeServerMessage(
+                ServerMessage.ModelList(
+                    models = swapper.availableModelIds(),
+                    current = swapper.currentModelId(),
+                ),
+            )))
+        }
+
         // Initial resource gauge so a freshly-connected client renders VRAM/RAM immediately.
         val snapshot = withContext(Dispatchers.IO) { module.resourceProbe.probe() }
         session.send(Frame.Text(ProtocolSerializer.encodeServerMessage(snapshot.toResourceStatus())))