Combine figures 3 and 4

[RobvanGastel]

I combined experiments 3 and 4 for issue #19.

Only 2 things remain unsolved. First, on line 84 the n_params is hardcoded to 3, should this be passed like in Figure 3?

rl-oed/examples/Figure_4_RT3D_chemostat/train_RT3D.py

Line 84 in 50cd5a9

size=(cfg.environment.n_parallel_experiments, 3)

Second, experiment 4 is not averaged on the experiment side, should we average on the experiment side or use weights and biases?

If these are resolved we can remove Figure_3 and Figure_4.

Changes

• Added a .gitignore file to ignore pycache directories
• Added argparse arguments to select config files

[RobvanGastel]

Set line 84 to n_params instead of hardcoding, and averaging using wandb grouping for figure 4, the ticket should be complete now.

[Johnny1188]

Good job! Just a few points:

• It seems like the merged script doesn't contain the latest version of the original two scripts (for example loading from a checkpoint is missing). There might have been some PRs on the master branch that your fork didn't mirror, could you please update it?
• Having multiple trials right there in the script is a good idea but I think the metrics and model checkpoints that get saved locally overwrite the metrics and checkpoints from the earlier trials (the save path cfg.save_path is still the same).

[RobvanGastel]

Good catch on the save_path for the trials, I will add the trial id to separate every output and also merge with the checkpoint on the master branch.