[Enhancement] Improving cluster state updates by concurrently executing appliers and listeners

[jainankitk]

Is your feature request related to a problem? Please describe

I am wondering if the ClusterStateAppliers and ClusterStateListeners need to be single threaded. While consuming single ClusterChangedEvent, we can execute all the appliers and listeners in parallel, given there is no implicit guarantee on the order of these appliers and listeners anyway. For example:

for (ClusterStateApplier applier : clusterStateAppliers) {
    logger.trace("calling [{}] with change to version [{}]", applier, clusterChangedEvent.state().version());
    try (TimingHandle ignored = stopWatch.timing("running applier [" + applier + "]")) {
        long applierStartTimeNS = System.nanoTime();
        // Execute all the appliers in parallel using separate threadpool
        applier.applyClusterState(clusterChangedEvent);
    }
// wait for all appliers to finish executing to prevent cluster state update thread to consume next event
// record latency for the cluster state applier operation
}

We can do same thing for listeners as well. Am I missing something?

Describe the solution you'd like

Execute all the appliers and listeners concurrently

Related component

Cluster Manager

Describe alternatives you've considered

No response

Additional context

No response