Since we want to follow the scikit-learn .fit type of execution for all our .run methods, the PatchscopeLogitLens and the ClassicLogitLens .run methods should take the same arguments.
This should be doable.
The SourceContext position attribute can be a sequence of integers. So, we can process more than one positions in one Patchscope "patch". This means, we can pass a sequence of integers to the PatchscopeLogitLens .run method as well.
We then simply need to change the ClassicLogitLens .run method to take the same arguments and adapt all scripts that use them.