Improved plotting
What does this MR do?:
- Rewrite the plotting script. Plots from the DL1 example config and plots from the umami example config
- The add_validation_file and add_test_file are now also used for both umami and DL1
- N_test number is now saved to the data frame when using evaluate_model.py. This way it does not have to be given in the plotting config
- Add memory growth so TensorFlow doesn't take all available GPU memory. In my tests, this reduced virtual memory usage from 30GB to 20GB without noticeable performance loss
- fix deprecation warning for matplotlib in PyATLASstyle.py
- Accuracy plot y label fixed
- made recommended tagger lines in epoch plots more distinguishable before - after