Restart simulation fixes

[MartinKarp]

Adds various arrays to the restart such that a restart no longer starts the time integration with time integration of order 1, but rather exactly progresses the simulation exactly as if the simulation did not restart but simply ran for longer. Should be backwards compatible, but could use further testing.

Additions to the checkpoint file:

• Lagged arrays for the scalar
• Lagged extrapolated contributions (abx,aby,abz etc.) for both the scalar and fluid
• Dt and tlag arrays

To be improved:

• checkpoint should be moved out of fluid and perhaps restructured entirely.
• each solver should add what it needs to restart into the checkpoint upon init, currently, I do this in the case file.
• Feel like I did a lot of code cuplication
• I did a test for debug message in the scalar solver, can we do this better?

To be noted:
There are two ways to restart in some sense, one is to save everything (like I do in this PR), and another is to recompute certain values from the field values at earlier stages (abx, aby, etc. can technically be calculated from the lagged u,v,w,s fields). However, due to the coupling with the forcing to calculate the abX arrays I chose to simply save everything. I feel this avoids a bunch of issues if someone would change the forcing or similar. We should just enforce that all new solvers add exactly what they need to the checkpoint upon init. The good thing about the recomputation part would be that one could restart easily from field files, but I think we can add some kind of reader for that as well in the future if it is necessary.

[njansson]

Great stuff! 🍻
I'll cleanup the rest and also backport it to release/0.7