refactor: Enable the subsolver(s) in a sequentially coupled solver to be sequentially coupled too

[frankfeifan]

If a multiphysics solver sequentially couples the subsolver A and subsolver B, the current architecture of CoupledSolver doesn't allow the subsolver A or B to be sequentially coupled. It is mainly because of limitations in two places:

1. setupSystem for subsolver(s). Now, there is no setupSystem well defined for the sequentially coupled (sub)solver.

The fix is to check if the subsolver is sequentially coupled as well. If so, we loop on the sub-subsolvers (if it is fully implicit) to set up the system.

2. nonlinearImplicitStep for subsolver(s). Now, there is no nonlinearImplicitStep well defined for the sequentially coupled (sub)solver.

The fix is re-format the sequentiallyCoupledSolverStep and add a method called sequentiallyNonlinearImplicitStep, which stores the original sequential loop codes. Also, we re-define the nonlinearImplicitStep for the coupled solver, and check if the subsolver is sequentially coupled. If so, we loop on the sub-subsolvers (if it is fully implicit) to perform nonlinearImplicitStep.