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.

test/errors.test.js

@@ -106,3 +106,20 @@ test('presentError can suppress toast when job drawer owns UI', (t) => {
   t.is(opts.toast, false)
   t.is(opts.tray, false)
 })
+
+test('RPC timeout is operation timeout not connection problem', (t) => {
+  const err = new Error('timeout of 30000ms exceeded')
+  err.code = 'TIMEOUT_EXCEEDED'
+  const info = explainError(err, 'removeContainer')
+  t.is(info.code, 'TIMEOUT_EXCEEDED')
+  t.ok(/timed out|timeout/i.test(info.title + info.message))
+  t.absent(/connection problem/i.test(info.title))
+  t.ok(/refresh|retry|docker/i.test(info.recovery))
+})
+
+test('timeout message without code is not connection problem', (t) => {
+  const err = new Error('timeout of 120000ms exceeded')
+  const info = explainError(err, 'removeContainer')
+  t.ok(/timed out|timeout/i.test(info.title))
+  t.absent(/connection problem/i.test(info.title))
+})