GitLab

Added read_local_nnz() which gives only the local data that is nonzero (for both sparse and dense).  Also added sparsify() which turns a sparse or dense tensor into a sparse tensor excluding all values that are zero or smaller than the threshold. sparsify() is used by read_local_nnz()

Accordingly, made read_local give all the zero values even when the tensor is sparse, with the goal of seemless compatibility/interchangeability of sparse and dense tensor codes