Experiment environment

[zabor2432]

We need to have mlflow for experiment tracking (tensorboard is also good, but mlflow should be easier to maintain)
The most robust solution is to have implemented solution no 4 from This URL before we will run final experiments the remote host don't have to be on for 24/7, it could any of our machines.

[zabor2432]

Czyszczę issues, skoro jest nas 3 na jedno issue, to można się podpiąć pod to zamiast mieć dwa (experiment pipeline i experiment environment)