Change order of preprocessing steps to follow Steve Luck's suggestions

[alexenge]

Prompted by Roman, who mentioned that re-referencing before ocular correction will project the eye artifacts at fontal channels into all the other ones

Here's the order of preprocessing steps as suggested in Steve Luck's book and website:

1. Filtering
2. Artifact correction (ideally first marking bad segment, but we don't do this [yet])
3. Re-referencing
4. Epoching
5. Artifact rejection
6. Some other stuff we don't need (e.g., averaging, difference waves)
7. Scoring (averaging across time and channel ROIs)

[alexenge]

One downside is that we still want to perform TFR analysis on unfiltered data, and so if we filter early, we need to apply all the other steps (OC, re-referencing, etc.) twice, once on the filtered and once on the unfiltered data