Solve flow directly on connectivity graphs with transmissibility values

[herve-gross]

Solve flow directly on connection graphs with transmissibility values
Provide connectivity and transmissibility information explicitly instead of computing these from the mesh directly. This allows for non-neighbor connections, and for disconnected neighbors.

This is a first step in the creation of the degree of freedom manager: instead of solving on the mesh points directly, we create an abstraction of the mesh based on the numerical scheme and supply solvers only the information they need for the target objects (any combination of volumes, faces, edges, or vertices).

Benchmarking with intersect and AD-GPRS
This approach would allow for a closer benchmarking opportunity with IX and AD-GPRS (these simulators can consume graph/transmissibillity).

The first step is to supply GEOSX with graphs/transmissibility grids.

[AntoineMazuyer]

I am digging up this issue as we have a use case involving a connectivity list and non conforming mesh.

Now that we have the DOF Manager, things may be easier.
I think the mesh can also be imported as it will hold some properties (porosity). The mesh will also be the support for volume computation and output... and I think it's all ? The idea is to bypass the TPFA stencil computation and directly take the one which is given (by eclipse for instance).

I can start investigating to see if there is any major issues with doing that.

If you have any thoughts/idea/comments about that, please feel free to share !