save/restore ThermoPhase state assumes density is an independent variable

[speth]

The functions Phase::saveState and Phase::restoreState assume that temperature and density are the independent thermodynamic variables, which is not true for several of Cantera's thermo models. For these models, the save/restore feature does not work correctly. For example (in Python):

import cantera as ct
phase = ct.ThermoPhase('test_problems/cathermo/issp/IdealSolidSolnPhaseExample.xml')
T0, p0 = phase.TP # T0 = 500, p0 == 101325
z = phase.state # calls Phase::saveState
phase.TP = 600, 2*ct.one_atm
phase.state = z # calls Phase::restoreState
assert phase.T == T0, 'temperature'
assert phase.P == p0, 'pressure'

Here, the pressure is left at 2 atm, despite the call to restoreState. In other cases, the restoreState call will fail because the derived class overrides the setDensity function to throw an exception.

Among other things, this error contributes to making these phases difficult to use with the Solution class in Python and Matlab, where the default behavior of instantiating a dummy Transport object triggers calls to saveState and restoreState, with the latter failing.

This problem can be resolved by making the saveState and restoreState methods virtual and having them work with the appropriate set of independent variables for each phase type.

[ThanasisMattas]

Hello! I would like to have a look on this issue. I think this could be handled parallel to this @todo entry commented in Phase.h:

@todo
- Make the concept of saving state vectors more general, so that it can
handle other cases where there are additional internal state variables,
such as the voltage, a potential energy, or a strain field.

So, possible overrides at derived classes will use their access methods for every other state variable, like voltage(). So, is just a virtual keyword enough here? And both methods can be overridden at the derived class. More specificly, I would like to ask if this requires any re-implementation of the overridden methods so that they handle the proper state variables (and, corresponding to the todo entry, to add any state variable to the state vector).