WIP: Outline of model integration + preprocessing

[minneker]

Work in progress. Preprocessing includes low-pass filtering, notch filtering and decimation. Model needs to dynamically change input size based on user inputs. I suggest we restrict user input for bin size so we can anticipate a couple of fixed input sizes. The reason this is tricky is because the architecture of EEGNet depends on the number of samples.

The main item necessary for integration is loading .edf file in the backend. It is unclear to me how to read a binary string as an .edf file (haven't found package functionality for that), and I'm thinking we may have to pass it to the backend another way since we rely heavily on mne in the backend.

[saketh-n]

Makes sense Rob. edf issue is resolved, as for samples and model architecture we could change the architecture to only take one to n samples and run many models in parallel

EEG_NET seems to take variable # of samples, but I'll look into it further