Create a higher rank dataset with dim 0 being a JOIN dimension. Join implies that the joined datasets occupy the same physical dimension, and this can be thought of as the "array of" index. Note DEPEND_0 is treated as a special case of join. Note this dataset is mutable, and clients should not mutate it once the reference is made public.
copy all the records into this JoinDataSet. Note this is a shallow copy, and changes to one of the element datasets is visible in both JoinDataSets. TODO: this is probably under implemented, for example element properties.