ENH: concept graph building functionality

[Sann5]

• Adds functionality so users can construct graphs on a subset of the cells in each image.
• Adds functionality to save attributes in the nodes of the generated graphs.
• One of the plotting functions is modified such that it can plot only subsets of the cells, thereby allowing visualization of subset graphs.
• As a result of these added functionalities the API of the athena.graph.build_graph() function has changed. The notebooks/tutorials were adjusted accordingly.
• Some old code was formatted to comply with flake8 and black. All new code complies with black and flake8.
• It also adds a test suite to test these added functionalities.

[adrianomartinelli]

Do you think I would make sense to include the graph concept training in the release? I think it would be nice to have the option to pip install athena or pip install "athena[gcl]" for which it installs dependencies for the deep learning with the graphs. In addition we can have a tutorial on how to train a basic model on the graphs. I think this is essential that people can use it.

[Sann5]

This is a great question. I think it does make sense to include it. The package (the functions and classes that are used in the workflow) is almost meaningless without the workflow. The task of generating graphs, attributing them, training several models, and comparing them is so large that it is only practical to be used with the workflow. Not to mention that leaving out the workflow would mean to let a lot of work go to waste.