Running the postprocessing if a solution is not fully converged can leave side-effects

[fw-bh]

Running the postprocessing on a not-converged solution leads to changes in connection, component or bus parameters in a way, that a result for a parameter may not be identical to the original input specification, i.e. isentropic efficiency was specified as 0.8 and ends up being 0.9. With this, a new solve call will impose the (wrong) result of that parameter to the model.

A reasonable solution would be to either

• run the postprocessing only, if a solution is fully converged (this however is not healing the original reason)
• or instead only update those component, connection and bus parameters, that are not inputs (i.e. where is_set is False). Those that are inputs should be stayed untouched.

The second option might work well together with the changes suggested in #591 as it can easily be automized and does not need to be changed for every single parameter individually.