Need to break storage out of sparse matrix assembler #213
Description
We're going to run into problems with dynamic problems on GPUs. Currently the AMDGPU and CUDA extensions use nzval storage in the matrix pattern. This is not duplicated for mass, damping, and hessian so we'll run into collisions.