Add problem set and solutions plus visual learning metadata

[matchings]

This PR adds the following things:

• The Visual Learning dataset (a.k.a. learning mFISH a.k.a. cell types and learning a.k.a. ophys-mFISH)
• A notebook in /metadata that generates the metadata .csv for the Visual Learning data by querying docdb (v2 schema)
• A problem set that is generalizeable to any dataset, walking students through replicating a classmate's figure then continuing to explore the features of the dataset as they compute PSTHs and signal/noise correlations
• Solutions versions of the problem set for the available datasets (V1DD, BCI, Visual Learning, Dynamic Routing)

All solutions notebooks run in full if the metadata .csvs for all datasets are present.

There may be some additional changes to the exact conditions plotted in the BCI notebook and the exact example used for DR along with any other requested changes based on feedback. But this should be ready to release as is.

Capsule is here: https://codeocean.allenneuraldynamics.org/capsule/6206290/tree

[leesuyee]

I'm not totally sure what's happening here, maybe updating the order in which these cells were run? I don't think you want to push this in the PR - can you discard the changes?

[leesuyee]

Same comment as above - can you revert the changes made to all the metadata notebooks except for visual_learning? I think this will create a merge conflict later on.

[leesuyee]

I think this code could be simplified. We should discuss further metadata fields to add to make these assets more searchable. Probably not enough time to do this, so we can push this through.

[leesuyee]

I think we could cut part 1 as this is redundant with the tutorials and the group exercise we'll ask them to do. The problem set is quite long and I don't think we'll need to structure them through how to explore a dataset for a 3rd time. We could keep that bit in the solutions, as it could be a useful reference for the students later.