MRG, DOC: Restore I/O content via new "tutorials"

[larsoner]

This PR:

1. Restores the (relevant) content from I/O that disappeared in MRG, DOC: Integrate manual parts #6767
2. Unifies our presentation of "how to read data from systems" by making two new "tutorials", one for MEG, one for EEG, that link to how to read data for each system (alongside relevant considerations). Eventually we can maybe actually have some code and plots for this. The tables for readers have gone away as a consequence, as they became redundant with the TOC at the top of these two pages.
3. Unifies how things are treated in implementation.rst. Previously some of the content was actually in the page, others were linked out with a toctree. Now it's all linked out in a toctree.

@agramfort there is a place in the EEG reading tutorial where the channel location stuff could live, but a dedicated/separate tutorial seems like a reasonable thing. We can easily link to it from plot_20_reading_eeg_data.py. Okay for you?

closes #6792

[codecov]

Codecov Report

Merging #6809 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #6809      +/-   ##
==========================================
- Coverage   89.63%   89.63%   -0.01%     
==========================================
  Files         422      422              
  Lines       76694    76715      +21     
  Branches    12549    12551       +2     
==========================================
+ Hits        68745    68763      +18     
- Misses       5129     5132       +3     
  Partials     2820     2820

[agramfort]

I added stuff on reading montages in #6764 this should be integrated with what we have here.

…

[larsoner]

Two big changes are:

1. New section here
2. This page becoming just a toctree rather than containing some of the content but not other parts

this should be integrated with what we have here.

Agreed, let's shoot to merge #6764 first, rebase this, then decide where things should live (probably a question for @drammock).

[agramfort]

we need @drammock <https://github.com/drammock>'s input here.

…

[drammock]

1. I think the two new tutorials should go in "intro" section rather than getting their own section.
2. (unrelated point) Probably the "background on projectors" should move to the preprocessing section
3. I'm not wild about the changes to implementation.rst. I would have gone the other direction --- making the algorithm content into its own page with a contents:: :local: instead of making everything into a :toctree:. With the TOC Tree, the revised "implementation details" page is overwhelmingly long, not an improvement in my opinion.

OK if I take over?

[larsoner]

From talking on Slack with @drammock we discussed 👍

1. keeping as a separate section to allow expansion to other types (fNIRS, ECoG, sEEG) in the future
2. @drammock doing whatever needs to be done :)
3. Making implementation.rst have the content itself instead of links to content pages

[agramfort]

@drammock you have carte blanche !

[larsoner]

Let's nuke this. It can live in the MNE-C manual. (In general I'm trying to get rid of any reference to a mne_* command and leave those to the MNE-C manual.)

[larsoner]

@drammock let me know if/when you are done. I'd like to:

1. Remove the mne_coil_def stuff
2. Add links from the single table to the individual ones, e.g. with meg-ctf, eeg-bv links. I had it done this way at one point before I restructured things, I think it will be nice for shortcuts from the big table to the tutorial sections.

[larsoner]

Changes:

• Remove overview/command_line.rst since we no longer linked to it from anywhere, and should be part of the MNE-C materials anyway
• Used git grep "\Wmne_\w*_\w\+" to find instances of the MNE-C commands being used. Eradicated the ones in the docs (left ones in the code that were refs to equivalent code snippets in C)