ENH: Write out raw data and SSP events

[larsoner]

Before merging …

• Changelog has been updated (docs/source/changes.md)

Closes #831

Also does some cleanup of processing= where we errantly used some uncleaned data -- not just no SSP (which would be taken into account during inversion at least), but no ICA, no regression, and no filtering it seems.

[larsoner]

All green, ready for review/merge from my end @hoechenberger @agramfort !

I pushed a [ci skip] commit just to remove an errant line from v1.6.md so no need to run CIs again, can just be merged without waiting for requirements if there are no review comments.

[larsoner]

... you can see an example of the Raw (clean) here:

https://output.circle-artifacts.com/output/job/5a50001f-886a-48c7-8e3f-cbcec122019e/artifacts/0/site/examples/ds000248_base/sub-01_task-audiovisual_report.html#Raw__clean___run__01_

I think it's useful enough to have these raw runs for comparison with the original and filtered ones in terms of time traces and PSDs to just have this always on (rather than add an option for it, which was a potential idea in #831)

[drammock]

@larsoner I took a look as requested. Changes look reasonable (with the caveat that, as you know, I'm not super familiar with this codebase yet). Just one comment.

[drammock]

this is the third time I've seen the same inline "filt-or-regress" if-else. Makes me worry about forgetting to copy it where it's needed, when adding new code. Not sure if there is a sensible way to reduce that cognitive overhead / duplication that wouldn't feel over-engineered, though.