Using a centralized way to handle path files (raw / prepared / clean and file names)

[antoan2]

This might be a environment file named .env.file_path loaded as an env file in the docker compose.
It would allow the download_datasets.sh script to download the data at the right place. We could also improve the main functions of the preparation scripts to get paths as arguments. Then, the scripts/prepare_data.py will be the only place where the environment is read and given as input of the preparation functions.

Once #9 is implemented, we will not have that much paths to save.