Extend plot_csd support to SEEG, ECoG, and DBS channel types

[Aniketsy]

Fixes #13711

[tsbinns]

Thanks for opening this, definitely on the right track! Some comments below.

[Aniketsy]

@tsbinns thanks for the review and for suggesting this approach, it makes much more sense than my earlier implementation. please have a look and let me know if any further improvements are needed.

[tsbinns]

Good changes! A couple small comments on the code, and a slightly bigger one for to refine the test. Let me know if anything's unclear.

[tsbinns]

@larsoner @wmvanvliet I noticed that the existing behaviour when plotting is that if no valid channel types are present (currently only EEG, Mag, Grad; this PR would change it to all valid data types), it just returns an empty figs container list.

I was a bit surprised by this. My expectation was at least a warning would be given that there is nothing valid to plot. WDYT?

[tsbinns]

Good progress! A couple minor comments to tidy things up.

[tsbinns]

@Aniketsy I'd say this is looking good, but I'd still like to get this comment addressed (#13713 (comment)) and maybe also a review from @wmvanvliet.

Please also add your name to doc/changes/names.inc and use the :newcontrib: role on your name in the changelog (no underscore is required).

[Aniketsy]

ERROR: Could not find a version that satisfies the requirement statsmodels>=0.15.0.dev697 (from versions: none)
ERROR: No matching distribution found for statsmodels>=0.15.0.dev697

I noticed this ci failure in two PR's @tsbinns should we relax the statsmodel version requirement?

[tsbinns]

That was addressed in #13728. Don't worry about merging the updated branch until there's new code to push. Will save us some cost on running CI jobs.

[Aniketsy]

Just saw it. Makes sense