Extra logging

[icunnyngham]

Environment specific logging which can be saved to CSVs or to wandb via redis when using DistribPolicyGradients.

Note:

• I put the worker-side aggregation step into the MARL agent because I needed access to cumulative timesteps right before experience is sent. If this could be achieved more cleanly, keeping extra_logger confined to DistribPolicyGradients, I would be happy to change this
• All the places where i had to add extra_logger as an argument for build_rlgym_from_config are pretty ugly, but I wasn't sure how else to achieve this. That function needs access to the logger to inject it into the rewards and obs functions.