Implement AROMA classifier

[tsalo]

Closes #5.
I took the metric and classifier code from ME-ICA/aroma for this.

Changes proposed in this pull request

• Incorporate pure Python implementation of AROMA feature calculation and classification.
• Create denoising workflow that supports "aggressive", "non-aggressive", and "orthogonalized aggressive" approaches.

[codecov]

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

Thanks for integrating Codecov - We've got you covered ☂️

[tsalo]

@effigies I'm thinking it's time for me to write a test or two of the workflow I'm adding here. I'm thinking of just running fMRIPrep on a single subject's data in "full" mode and uploading the relevant derivatives to Box to use as test data. Do you have any recommendations for OpenNeuro datasets to grab from?

EDIT: I'm going to go with 28andHe since I'm planning to play with it at some point anyway.

[tsalo]

There are probably a lot of things I can do to improve the AROMA outputs (e.g., I ended up moving the component metadata, like variance explained, into a TSV that could maybe be moved back into the JSON), but I think the only thing I need to do to actually get the AROMA workflow running is modify the DerivativesDataSink to support the ICA outputs.

[tsalo]

I don't know what's happening with the mock_config, but everything else looks good to me.

[tsalo]

We can use https://gin.g-node.org/tsalo/ds005115-fmriprep-test as the base test data for the AROMA and denoising workflow, since it's a small dataset (50 volumes). Just need to write up a job to grab the data and a test to run the pipeline.

[tsalo]

Merging now.