Fix Overly Optimistic Request Deduplication

On master failover we have to resent all the shard failed messages,
but the transport requests remain the same in the eyes of `equals`.
If the master failover is registered and the requests to the new master
are sent before all the callbacks have executed and the request to the
old master removed from the deduplicator then the requuests to the new
master will incorrectly fail and the snapshot get stuck.

Closes elastic#51253

server/src/main/java/org/elasticsearch/snapshots/SnapshotShardsService.java

@@ -358,6 +358,9 @@ private void syncShardStatsOnNewMaster(ClusterChangedEvent event) {
             return;
         }
 
+        // Clear request deduplicator since we need to send all requests that were potentially not handled by the previous
+        // master again
+        remoteFailedRequestDeduplicator.clear();
         for (SnapshotsInProgress.Entry snapshot : snapshotsInProgress.entries()) {
             if (snapshot.state() == State.STARTED || snapshot.state() == State.ABORTED) {
                 Map<ShardId, IndexShardSnapshotStatus> localShards = currentSnapshotShards(snapshot.snapshot());

server/src/main/java/org/elasticsearch/transport/TransportRequestDeduplicator.java

@@ -53,6 +53,14 @@ public void executeOnce(T request, ActionListener<Void> listener, BiConsumer<T,
         }
     }
 
+    /**
+     * Remove all tracked requests from this instance so that the first time {@link #executeOnce} is invoked with any request it triggers
+     * an actual request execution. Use this e.g. for requests to master that need to be sent again on master failover.
+     */
+    public void clear() {
+        requests.clear();
+    }
+
     public int size() {
         return requests.size();
     }