MSFR core multiphysics model improvements

[GiudGiud]

We're leaving a lot of performance behind right now in those MSR sims. A few to consider:

• [x] initialize the simulations from exodus rather than from a multiapp
• [x] better time stepping scheme for the transient
• [x] known-to-converge large time steps for the initialization
• [x] only do the viscosity rampdown at initalization
• [ ] separate precursor advection and fluid advection to avoid paying for a fully coupled solve