Navier-Stokes problem for shared memory AMGCL benchmarks

The system matrix and the right hand side used in the shared memory Navier-Stokes benchmark at https://github.com/ddemidov/amgcl_benchmarks.

This is an incompressible Navier-Stokes problem discretized on a non uniform 3D mesh with a finite element method. The overall system matrix for the problem was assembled in the [Kratos](http://www.cimne.com/kratos/) multi-physics package developed in CIMNE, Barcelona. The matrix has block structure with blocks of 4-by-4 elements, and contains 713'456 unknowns and 41'277'920 nonzeros.