Implement addBatchInDisk for big trees, which does not puts the tree in memory, and works directly over the db data, parallelizing for each CPU.