iMC changes outside of storm-pars

[linusheck]

Draft PR containing all changes I made for iMC algorithms that are outside of storm-pars. Tests in the storm-pars part (no PR for this yet) cover most lines in this PR.

In more detail, this PR contains:

• a new function for FlexibleSparseMatrix that adds new rows
• a SchedulerTrackingHelper for iMCs and a interval (=robust) scheduler index in the minmax solvers for warm-starting
• interval support for the (Iterative)MinMaxLinearEquationSolver
• an addNotCurrentModel function for SMT solvers that adds the constraint that the current model is not a solution
• AGNUM support for Z3
• interval MC end component elimination (RobustMaximalEndComponentDecomposition, RobustMaximalEndComponent)
• optimisations in ValueIterationOperator for iMCs
• a couple more functions for storm::Interval

[sjunges]

Does this depend on #588? That PR is still not ready for merging, is it?

[glatteis]

Does this depend on #588? That PR is still not ready for merging, is it?

It doesn’t depend on that PR I‘m pretty sure, because that PR (mainly) changes things in storm-pars. The open discussion was about the behavior of the CLI flag but that is definitely in storm-pars.

[glatteis]

Ok, I see that there are indeed some changes outside of storm-pars in there, but those are not under discussion :D

[tquatmann]

Thanks for the great work!

I think some test cases for the robust mec decomposition and the scheduler tracking for robust value iteration would help with maintaining the new features.

[linusheck]

If we agree to postpone these last two things (we've agreed on the end component decomposition refactoring but not yet on the SMT solver interfaces), we can merge this PR. Currently, the only place where SMT results are rounded are after big-step preprocessing. The changes to fix the SMT solver APIs are a lot of source code changes, but very doable later. See issue #713.

[linusheck]

seems that spot is failing :(

[volkm]

One issue with Spot will be fixed by #703, the other ones are tracked in #707. Other than that your CI seems to be successful.
We should merge it after version 1.10 is released.