[Analytics] Support data partitioning for multiple clusters coordinated write

In the coordinated write, data partitioning should consider the consolidated ring topology from all write-target clusters. Thus, the produced (spark) partitions do not span across multiple nodes, causing inefficiency.