ENH: Channel tags

It would be nice to be able to tag channels, for example as belonging to a particular "group".

• for MEGIN MEG, this could be 1) each triplet of sensors that share a common spatial location, and also 2) which selection they are a part of; e.g. MEG0111 could have tags ('011X', 'Left', 'temporal') or so (part of the 0111/2/3 triplet, in the Left-temporal group in .sel)
• for iEEG, this would probably be sEEG shafts / ECoG grid
• for OPMs, manufacturers creating one "sensor" with multiple sensitive volumes (even if spatially offset from one another) will probably want these groups

I propose we add to our new enhanced channel block (https://github.com/mne-tools/fiff-constants/pull/33) the ability to specify which group each channel belongs to. I think it makes sense to add a new entry like:

ch_tags		352	str	-	"Channel tags"

@agramfort @alexrockhill @jasmainak thoughts? I think this will allow us to make visualization functions automagically work better, for example by:

1. channel groups each get their own color from the mpl color cycle in plot_alignment et al.
2. mne.channels.find_layout could construct linear layouts on a per-channel-group basis (e.g., for sEEG shafts) for use with mne.viz.plot_topo

If this makes sense, I'll open a parallel issue over in fiff-constants. But I figured we could think about and discuss it a bit here first!

Start by reviewing the proposed enhanced channel block and the referenced fiff-constants pull request. Consider how a ch_tags entry would interact with plot_alignment and mne.channels.find_layout; done means the channel-group design and follow-up scope are agreed.