Fix (and test) for a deadlock that can happen while waiting for protocol info (#12633)

# Description

The local socket PR introduced a `Waitable` type, which could either
hold a value or be waited on until a value is available. Unlike a
channel, it would always return that value once set.

However, one issue with this design was that there was no way to detect
whether a value would ever be written. This splits the writer into a
different type `WaitableMut`, so that when it is dropped, waiting
threads can fail (because they'll never get a value).

# Tests + Formatting

A test has been added to `stress_internals` to make sure this fails in
the right way.

- 🟢 `toolkit fmt`
- 🟢 `toolkit clippy`
- 🟢 `toolkit test`
- 🟢 `toolkit test stdlib`

tests/plugins/stress_internals.rs

@@ -1,3 +1,5 @@
+use std::{sync::mpsc, time::Duration};
+
 use nu_test_support::nu_with_plugins;
 
 fn ensure_stress_env_vars_unset() {
@@ -75,6 +77,30 @@ fn test_failing_local_socket_fallback() {
     assert!(result.out.contains("local_socket_path: None"));
 }
 
+#[test]
+fn test_exit_before_hello_stdio() {
+    ensure_stress_env_vars_unset();
+    // This can deadlock if not handled properly, so we try several times and timeout
+    for _ in 0..5 {
+        let (tx, rx) = mpsc::channel();
+        std::thread::spawn(move || {
+            let result = nu_with_plugins!(
+                cwd: ".",
+                envs: vec![
+                    ("STRESS_EXIT_BEFORE_HELLO", "1"),
+                ],
+                plugin: ("nu_plugin_stress_internals"),
+                "stress_internals"
+            );
+            let _ = tx.send(result);
+        });
+        let result = rx
+            .recv_timeout(Duration::from_secs(15))
+            .expect("timed out. probably a deadlock");
+        assert!(!result.status.success());
+    }
+}
+
 #[test]
 fn test_exit_early_stdio() {
     ensure_stress_env_vars_unset();