[BUG] Stuck segments upload leads to high refresh lag

[ashking94]

Describe the bug
There is a use case where when segments are getting uploaded to remote store and a flush/force merge happens, there is a possibility that the segment tracker's data can get erased while there is async refresh retry is happening. This leads to latch not getting counted down.

OpenSearch/server/src/main/java/org/opensearch/index/shard/RemoteStoreRefreshListener.java

Lines 388 to 403 in a2febe9

	for (String src : filteredFiles) {
	// Initializing listener here to ensure that the stats increment operations are thread-safe
	UploadListener statsListener = createUploadListener();
	ActionListener<Void> aggregatedListener = ActionListener.wrap(resp -> {
	statsListener.onSuccess(src);
	batchUploadListener.onResponse(resp);
	}, ex -> {
	logger.warn(() -> new ParameterizedMessage("Exception: [{}] while uploading segment files", ex), ex);
	if (ex instanceof CorruptIndexException) {
	indexShard.failShard(ex.getMessage(), ex);
	}
	statsListener.onFailure(src);
	batchUploadListener.onFailure(ex);
	});
	statsListener.beforeUpload(src);
	remoteDirectory.copyFrom(storeDirectory, src, IOContext.DEFAULT, aggregatedListener);

To Reproduce
Run indexing for a long time.

Expected behavior
Refreshes should never get stuck.

[ashking94]

This is fixed with #11896