Ported to pytorch 2.0 and lightning 2.0

[HansBambel]

This is a big refactor of the (kind of outdated) code.

The dependencies for almost all packages were updated. Most importantly, pytorch is now 2.0.0 and lightning uses 2.0.

I have not fully trained the models to check whether everything works correctly, but it is possible to execute the train_precip_lightning.py script.

~test_precip_lightning.py was not tested yet.~

Next things to do:

• [ ] Fully train a model
• [x] Run test_precip_lightning.py
• [x] Runtrain_SmaAtUNet.py