Notify replaced plugin processes at upgrade grace start

Send rpc.superseded with the newly active hash before activation returns.
Add an opt-in SDK hook and drain helper that closes SDK listeners and idle
connections without cancelling active handlers or background work.
Document the additive notice and preserve existing grace, stop and restart rules.

Cover immediate notices, ignored notices, early service restarts, SDK draining
and duplicate-hook diagnostics.

Card: SAU-352

cli/src/upgrades.rs

@@ -12,6 +12,7 @@ pub(crate) struct OldProcess {
 pub(crate) struct Upgrade {
     plugin: String,
     previous: String,
+    current: String,
     grace: Duration,
     processes: Vec<OldProcess>,
     services: Vec<(Service, Option<watch::Receiver<bool>>)>,
@@ -67,6 +68,7 @@ impl Host {
         Upgrade {
             plugin: plugin.into(),
             previous,
+            current: current.into(),
             grace: Duration::from_secs_f64(grace),
             processes,
             services,
@@ -74,6 +76,13 @@ impl Host {
     }
     /// Lets old calls finish, stops them after grace, and starts each lifecycle hook once.
     pub(crate) fn begin_upgrade(&self, upgrade: Upgrade) {
+        // Queue the notice before returning from activation, without waiting on guest code.
+        for process in &upgrade.processes {
+            self.send(
+                &process.pid,
+                json!({"jsonrpc":"2.0","method":"rpc.superseded","params":{"current":upgrade.current}}),
+            );
+        }
         let host = self.clone();
         tokio::spawn(async move {
             if host.0.shutdown.load(Ordering::SeqCst) {