treating memory leak in thickness parameter study

[dion-w]

looking into #1442 i also had some problems with memory usage, especially on a cluster with creating the solver, simulation and solution object multiple times in a loop.
i also saw that it is recommended to setup the solver and simulation object outside of the loop to get rid of memory leak.

For my parameter study though i change the thickness of the electrode every loop and therefore (i think) the model needs to be discretized every loop, so the simulation object can't be created outside the loop.
All the other parameters can be made input parameters, i guess, but not so for the thickness i think?

What would you recommend in this setup?