Creating a large dask array backed by scipy.sparse.csr_matrix

Question

I have a function, call it get_block(I, J), which takes two index sets, I and J, and returns a scipy.sparse.csr_matrix wi