Initialization step for physics solvers

Describe the issue
We currently have a few different functions that can be used for initialization of the physics solvers. The role of each one of them is not always clear so physics solvers tend to use them a bit inconsistently.

The initialization functions are:

• initialize_postMeshGeneration (I am not sure this is still used)
• initializePreSubGroups
• initializePostSubGroups
• initializePostInitialConditionsPreSubGroups
• initializePostInitialConditionsPostSubGroups

Proposed cleanup
I propose to define what these functions do for a physics solver in SolverBase and have a final implementation there and then have physics solver specific functions to initialize things. Something like

void initialize_postMeshGeneration override final;
void initializePreSubGroups  override final;
void initializePostSubGroups  override final;
void initializePostInitialConditionsPreSubGroups  override final;
void initializePostInitialConditionsPostSubGroups  override final
{
  initializePrimaryVariables();
  initializeConstitutiveModels();
}

virtual void  initializePrimaryVariables();
virtual void  initializeConstitutiveModels();

This way we could avoid patterns which seem to be a bit arbitrary like the fact that this step is performed here https://github.com/GEOSX/GEOSX/blob/00dd40335a8b3dae72adb5ebc4e7b181e7b62051/src/coreComponents/physicsSolvers/fluidFlow/SinglePhaseBase.cpp#L231 but other initializations are performed here https://github.com/GEOSX/GEOSX/blob/00dd40335a8b3dae72adb5ebc4e7b181e7b62051/src/coreComponents/physicsSolvers/fluidFlow/SinglePhaseBase.cpp#L339 instead. Similarly, the linear algebra parameters for the physics solvers are not always set in the same function. These small inconsistencies make it hard when writing a coupled solver because the order in which things occur is not always clear.

I guess we could also introduce a new intialization step, less generic than the one provided by Group, which, for example, explicitly takes DomainPartition as an input.

Start by reading SolverBase and the initialization hooks listed in the issue, then compare the two referenced sections of src/coreComponents/physicsSolvers/fluidFlow/SinglePhaseBase.cpp. Trace when each initialization function and the linear algebra setup runs. Done means the lifecycle is explicitly defined and physics-solver-specific initialization has a consistent, documented order.