Fix container remove timeouts and make force-remove reliable.

Lifecycle RPCs now use a 120s operation timeout, remove cleans up
stats/terminal/log streams then SIGKILLs before force-remove, and the
UI awaits the request instead of a fragile 30s wait race.

server/services/stats.js

@@ -266,6 +266,26 @@ function destroyStatsEntry(id) {
   statsCache.delete(id)
 }
 
+/**
+ * Drop live stats stream for a container before stop/remove so Docker is not
+ * held open by NDJSON stats attachments (can delay force-remove).
+ * @param {string} id
+ */
+export function destroyStatsForContainer(id) {
+  if (!id) return
+  const needle = String(id)
+  destroyStatsEntry(needle)
+  for (const key of Object.keys(containerStats)) {
+    if (key === needle) continue
+    if (
+      (needle.length >= 12 && key.startsWith(needle)) ||
+      (key.length >= 12 && needle.startsWith(key.slice(0, Math.min(12, key.length))))
+    ) {
+      destroyStatsEntry(key)
+    }
+  }
+}
+
 async function collectContainerStats() {
   // Running only for streams; we still list all so stopped ids are cleaned up
   const running = await docker.listContainers({ all: false })