Integration/interoperability with autoreject

[christian-oreilly]

I just want to open the discussion on the possibility either to integrate autoreject (https://github.com/autoreject/autoreject) or build-in some kind of interoperability with that package. This may require figuring out first how we would integrate additional, optional steps, as starting to be discussed in #73.

[scott-huberty]

I think I'm in +1.

For one of my projects I actually preferred the RANSAC algorithm (which is available in autoreject; On that dataset, I was happier with the RANSAC bad channel detection than the Pylossless algorithm.. But Maybe it was just that particular dataset).

What I ended up doing was 1) Execute a Robust Average Reference and RANSAC via pyprep on my raw object 2) then running pylossless with run_with_raw 3) Tell the RejectionPolicy NOT to remove the channels flagged as Noisy (which would correspond to channels flagged by the pylossless algorithm).