Tpetra: parallel kernels for CRS matrix addition (#1367)
Created by: brian-kelley
-Makes TpetraExt::MatrixMatrix::add very thread scalable in most cases -Kernels using Kokkos to locally add two Tpetra CrsMatrices -One kernel for the sorted graph, matching col map case, and a more general one for unsorted graph and/or mismatched col map case -Unit tests for both kernels