fix: flush stderr after listening log and don't kill child on disconnect

Two fixes for jumpstarter-exec test failures:

1. Flush stderr after logging 'listening' so tests that poll for socket
   creation see the buffered log output.

2. Don't send SIGTERM to child process on client disconnect. Let child
   finish naturally so output forwarding threads can read any remaining
   stdout/stderr. Main thread reaps with wait().

Fixes test e2e_debug_json_logs_commands_and_io and related E2E tests.

Author: bennyz

rust/jumpstarter-exec/src/server.rs

@@ -137,6 +137,9 @@ pub fn serve_with(socket_path: &str, opts: ServeOptions) -> std::io::Result<()>
         "listening",
         &[("socket", json!(socket_path)), ("debug", json!(opts.debug))],
     );
+    // Ensure listening log is flushed so tests that poll the socket get
+    // visible evidence that the server is ready.
+    let _ = std::io::stderr().flush();
 
     while !state.is_shutdown() {
         match listener.accept() {
@@ -366,9 +369,8 @@ fn handle_exec(
                 _ => {}
             }
         }
-        if !reaped_ref.load(Ordering::Acquire) {
-            unsafe { kill(pid as i32, 15) }; // SIGTERM on client disconnect
-        }
+        // Don't kill child on client disconnect; let it finish naturally.
+        // The main thread will reap it with wait() after forwarding threads exit.
     });
 
     let status = child.wait()?;