[AGIPD][DARK] FIX: Select trains within `characterize_module()` and remove trains reduction nb cell

Fix reducing number of trains for all runs based on max(max_trains_list) after using require_all

Description

There was a bug that lead to not process all trains for all 3 runs. The runs were reduced to the maximum number of trains with images before removing the empty trains in the first place.

There are different changes in this MR surrounding the fix.

1. Only use max_trains_list for previewing information about the data to be processed.
2. Remove the trains with empty frames withing the processing function.
3. As the change were in a cell calculating the possible trains based on the available memory. Remove this cell as the calculation for the possible trains didn't make sense and we should in a different MR think about splitting the train processing.
4. To run the notebook interactively for 16 modules with about 600 trains each. Reduce the max number of multiprocesses to 6 instead of 12. The change was done after testing it with SPB Fixed and Adaptive reference runs in https://git.xfel.eu/calibration/planning/-/issues/70.

How Has This Been Tested?

Running SPB_DET_AGIPD1M-1 runs interactively and run automated test for all AGIPD DARK reference runs.

https://git.xfel.eu/detectors/pycalibration/-/merge_requests/610

https://git.xfel.eu/detectors/pycalibration/-/blob/02fda656d2f59b83ef5951c3da9fe093863f7822/tests/callab_tests.yaml

Relevant Documents (optional)

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist:

Reviewers

@calibration