removed network parameter, toar_data loads from both v2 and v1 and combines data, data is now in local time (without DST), data handler now raises error when no data intersection is available

revert many changes as multiprocessing approach will not work on HPC system with really big data due to memory issues

additional netcdf4 version install removed for setup_venv_hdfml.sh and reactivated ml Jupyter/... for mlt_modules_juwels.sh

reduced requirements_JUWELS_additionals to only not available (after mlt_modules_juwels are loaded), removed duplicated 'pip install -r' and uncommented matplotlib==3.2.0 and pandas==1.0.1 installations

changes in module versions to be compatible with HPC system (maybe diff pip version causes errors?) - put tensorflow at first

packages in *_additionals updated to new TensorFlow 2.5.0 environment and post_processing evaluate_generator(...) changed to evaluate(...)

changed to tensorflow==2.3.1 in requirements.txt and deleted requirements_tf2.txt and requirements_gpu.txt, because they are not needed anymore. Also updated mlt_modules_hdfml.sh

model.evaluate_generator changed, because it is deprecated and works with model.evaluate(generator, ...), updated mlt_modules_juwels.sh to Stages/2020 modules with Python/3.8.5

updated all run scripts with new Workflow class, HPC_setup script is calling now the new run_HPC.py script