Introduce full tensor permeability

[CusiniM]

@castelletto1 should be usable but I have not tested it yet.

[castelletto1]

I confirm it works. I ran an incompressible single-phase flow quarter five spot problem without gravity on two domains. The first case, the domain is aligned with the Cartesian axes and the permeability tensor is diagonal. In the second case, the domain is obtained by rotating the aligned domain and consistently updating the permeability tensor, which becomes full tensor. In both cases the grid is K-orthogonal and TPFA is correct. The computed Jacobian matrices are the same. Below two snapshots with details for the same cell (Id 52) in both domains.

I temporarily added xml and vtk files to reproduce the results in ./inputFiles/singlePhaseFlow/fullTensorPermeability We may end up adding a new smoke test.

@karimifard, @jafranc, @francoishamon: have a look and share if you think this is headed in the correct direction.

[jafranc]

I confirm it works. I ran an incompressible single-phase flow quarter five spot problem without gravity on two domains. The first case, the domain is aligned with the Cartesian axes and the permeability tensor is diagonal. In the second case, the domain is obtained by rotating the aligned domain and consistently updating the permeability tensor, which becomes full tensor. In both cases the grid is K-orthogonal and TPFA is correct. The computed Jacobian matrices are the same. Below two snapshots with details for the same cell (Id 52) in both domains.

I temporarily added xml and vtk files to reproduce the results in ./inputFiles/singlePhaseFlow/fullTensorPermeability We may end up adding a new smoke test.

@karimifard, @jafranc, @francoishamon: have a look and share if you think this is headed in the correct direction.

I let @karimifard and @francoishamon, who are more expert than I am, tell if it is what they had in mind. It looks like a right start and not adding too much complexity. There will be some bookkeeping traps along the road though and I like @CusiniM 's idea to have one tensor for the diag and one for off-diag so the full tensor case is use as the base and off-diag zeroed out by default.