wip: data+project management workflow

[bpinsard]

Here is a suggested WIP workflow and datasets/projects organization.
I tried to include examples to ground these ideas in concreteness.
Very preliminary, but I think the discussion is more efficient here than in meetings.

[hyruuk]

This looks great overall !
Some comments :

• Analysis : I think it would make sense that these would be linked to project repo (datapapers, regular papers etc...) instead of dataset repositories. For example, mario.training_dependent_rsa should be mario_learning.training_dependent_rsa. The reason I suggest this is that some analysis will involve several datasets, or might even be applied on different datasets (in an ideal world).

• I think we might want to enforce or encourage the use of invoke to assemble workflows when the repos are ready for reviews. This will allow users to easily use/reproduce the repo, as well as make sure that everything can be pipelined easily down to jupyter-book preprints.

• Paper names are often decided upon at the end of the process, while the paper repo might be created before that. Maybe we could keep the project name and add .paper behind ? E.g. mario_learning.paper