[Optimization] TaskBatcher to use ConcurrentHashMap for batching tasks per executor

[shwetathareja]

Describe the bug

In TaskBatcher.java, for batching it maintains a HashMap tasksPerBatchingKey and uses synchronized block to find duplicates using the batching key (which is based on task executor) and update the tasks for the given key. It can use ConcurrentHashMap instead which will only lock the segment and use computeIfPresent for computing updated tasks.

OpenSearch/server/src/main/java/org/opensearch/cluster/service/TaskBatcher.java

Lines 97 to 100 in 082f059

	synchronized (tasksPerBatchingKey) {
	LinkedHashSet<BatchedTask> existingTasks = tasksPerBatchingKey.computeIfAbsent(
	firstTask.batchingKey,
	k -> new LinkedHashSet<>(tasks.size())

Expected behavior
The tasks belonging to different batchingKey can be submitted in parallel and will make submitTasks faster.

[amkhar]

This is closed as part of #5827