[BUG] non-independent scenarios when batch scenario fails

[mgovers]

Describe the bug

A failing batch scenario may affect consecutive batch scenario

To Reproduce

• Create valid input data
• Create update batch with 2 scenarios
  • Ensure that the first scenario fails during calculation (e.g. sparse matrix error)
  • Ensure that the second scenario passes (e.g. by running it separately)
• Create a model with the input data
• Run powerflow calculations with the update batch with treading=-1
• Observe the result
  • The first scenario fails with the correct batch error
  • The second scenario has a different output than if it is run separately as an individual scenario (e.g. in a batch without the first scenario)

Expected behavior

• The first scenario fails with the correct batch error
• The second scenario produces the same output

Additional context

• The restore functionality of the main model at the end of a scenario is only called when a scenario passes. This results in a drift of the state of the main model. We do not have validation cases for that yet. The offending lines are https://github.com/PowerGridModel/power-grid-model/blob/main/power_grid_model_c/power_grid_model/include/power_grid_model/main_model.hpp#L472-L475
• The erroneous behavior was first observed in v1.6.44 due to a change in the restore behavior. This change was rolled back to the previous (also erroneous) behavior in v1.6.45, which has the same behavior as v1.6.43 and previous. See also Revert "Feature/remove needless id lookup" #459 .
• Threading might isolate some scenarios from others, and therefore might result in different outputs.

c.c. @petersalemink95 @TonyXiang8787 This is quite severe. I'm surprised no-one has run into this issue yet.