FV: Implement method for computing solution values on boundary faces

[lindsayad]

Reason

Right now a user has two choices: apply a Dirichlet condition or get implicit application of a zero normal-gradient condition. The latter is fundamentally wrong for physics like no-slip flows in which the pressure continually drops in the direction of flow. The former works ok for fully-developed flow with a single outflow boundary. However, in the future @GiudGiud will have multiple channel outflows and its nonsensical to try and apply multiple dirichlet pressure outflow conditions. Additionally, inlet physics will be slightly off, even for single channel cases, because the implicit assumption of zero normal pressure gradient at the inlet affects the solution. This can be seen in the no-slip diverging diverging channel test contained within #15644. The inlet mass flow rate should be equal to pi, and this is what is computed for free slip because the assumption of zero pressure normal gradient at the inlet is valid in that case. However, for no-slip the computed inlet mass flow rate is equal to 3.161216, which is slightly off due to affect of the invalid zero normal pressure gradient assumption.

So in summary: How do we compute the solution at boundary faces?

Design

We should research what projects like OpenFOAM do for this. Perhaps we will need some higher order construction of the solution.

Impact

Implement what intuitively to me feels like a very fundamental FV capability. For flow physics, this should allow us to implement correct inlet and outlet boundary conditions.

As a slight digression, we have two ways of applying outflow boundary conditions in our CGFEM Navier-Stokes implementation, both of which make more sense to me than what we can currently do for FV:

• NoBCBC: essentially just say flux = flux. This would lead to a non-unique solution for the strong form of the equations, but is meaningful for weak forms
• Natural boundary condition: normal traction force is zero

Neither of these outflow conditions constrain the pressure in anyway. Instead the only thing constraining the pressure is the mass conservation equation, which is exactly how it should be for the incompressible Navier-Stokes equations in which the pressure is just a Lagrange Multiplier ensuring that mass conservation is satisfied. (Similarly setting a Dirichlet condition for inlet velocity in CGFEM also does not constrain the pressure.)