Adding transient solver

[wdhawkins]

This PR adds TransientSolver. It is very loosely based on TransientKEigenSolver by @zachhardy.

TransientSolver has two modes controlled by the initial_state parameter: existing and zero. In existing mode, the transient solver evolves the solution of a previous solver. For example, to evolve the solution of a k-eigenvalue problem, one might do:

phys = DiscreteOrdinatesProblem(...)
keigen = PowerIterationKEigenSolver(problem=phys)
keigen.Initialize()
keigen.Execute()
tsolver = TransientSolver(problem=phys, initial_state="existing")
tsolver.Initialize()
tsolver.Execute() (or using Advance with a custom time stepping loop)

or, to evolve the solution of a steady-state problem:

phys = DiscreteOrdinatesProblem(...)
steady = SteadyStateSourceSolver(problem=phys)
steady.Initialize()
steady.Execute()
tsolver = TransientSolver(problem=phys, dt=0.001, theta=1.0, stop_time=1.0, initial_state="existing")
tsolver.Initialize()
tsolver.Execute()

In zero mode, the transient solver starts from a zero initial state and evolves the solution:

phys = DiscreteOrdinatesProblem(...)
solver = TransientSolver(problem=phys, dt=0.001, theta=1.0, stop_time=1.0, initial_state="zero")
solver.Initialize()
solver.Execute()

A number of examples are given in the transport_transient regression input folder.

This is a bit of a large commit. I tried to make it as small as possible by committing the time-dependent work in other commits, but it's still quite large.

Resolves #57
Resolves #58

[andrsd]

lbs problem-> discrete ordinates problem (ctor downcast to DiscreteOrdinatesProblem)