'Allow fixed composition/temperature on outflow boundaries' does not work correctly in time step zero, iteration 0 with open boundary

[jdannberg]

In time step zero, iteration zero, we do not have a solution for the velocity, so it is assumed to be zero for the purpose of determining which boundaries have inflow/outflow. #5481 fixes this for models with prescribed velocity boundary conditions, but not for open boundaries. With #5481, we will assume that all of the boundaries with zero flow are inflow boundaries and prescribe the temperature/composition. But it is actually not clear what the right solution should be in this case. One could argue that time step zero has a length of zero anyway, so it is better to not apply the boundary conditions because the material is not moving anyway, so at least we do not add any temperature/composition anomaly at a boundary that would otherwise have outflow (which can lead to oscillations later on if there is a large difference between the boundary condition and the initial condition, similar to what I described in #5042). But in many cases, the user may want to prescribe boundary conditions in time step zero, for example for instantaneous models and in particular at open boundaries where inflow is expected.

Therefore, I think it would be useful to have an input parameter that allows the user to either select if boundary conditions should be applied in step 0 (or not), or, a;ternatively if zero flow should be interpreted as inflow or outflow.