Clean up of public interface of SolverBase

@rrsettgast, @francoishamon, @castelletto1, @TotoGaz , @joshua-white, follwoing the infrastructure meeting, I have created a list of a few things that could be improved in the physics solvers interface. This issue expands what was already mentioned in https://github.com/GEOSX/GEOSX/issues/1653.

I have broken this down into a few different issues and I propose to use this one to keep track of all possible improvements.

1. Revisit the initialization step (https://github.com/GEOSX/GEOSX/issues/2178).

2. Centralize implementation of solverStep and account for sequentail solvers (https://github.com/GEOSX/GEOSX/issues/2176)

3. Functions specific to some physics solverts should be private but are not. This is, obviously, an easy thing to fix.

4. Some virtual functions in SolverBase probably don't need to be virtual. I think we could check if we really need all solvers to be allowed to override certain functions (e.g., solveLinearSystem, nonlinearImplicitStep, lineSearch).

Start by reading the SolverBase interface and the linked issues 2178 and 2176, then inspect the public physics-specific functions and virtual methods named here. The issue provides no file paths or agreed final API; done would require a defined and coordinated cleanup across the solver implementations.