Pipeline Meeting.

Notes from our meetings between MATLAB and pyLossless Devs to confirm that we pipeline steps are conceptually in agreement.

Note:

For every step. the first metric is on time.

Flag outlier chs

Flag outlier chs should be used in the re-reference function and we should not create a flag based off the output.

The reason is that we don't want to officially mark any channels before the first rereferencing to average. For example, on systems like EGI, the mono-polar reference (i.e. online reference to Cz) makes it so that channels close to Cz have a small voltage variance and channels far away (like skirt channels) have substantially larger voltage variance. So we want to put all the channels on the same playing field before making flags, but we don't want to include "comically" bad channels in the re-reference.