WIP: Add tsDSS

[larsoner]

@drammock FYI I'm working on implementing tsDSS. My plan is:

• Reformulate existing DSS code to use the same nomenclature from the AdC/Simon paper (@drammock feel free to look/confirm I did this properly so far)
• Reformulate to prefer mne.compute_whitener for the whitening step, since it's careful about channel scalings, projectors, etc. (and just using _pca will not be)
• Change the example to use the filtered data (faster)
• Extend the DSS calculation to be time-shift capable (tsDSS)
• Add simulation/array example from the paper

[coveralls]

Coverage remained the same at ?% when pulling 82f915d on larsoner:tsdss into 6440770 on mne-tools:master.

[drammock]

FYI, I am not very confident that my original implementation was sound --- I ended up rewriting this separately for the 2018 paper here. But it looks like you've re-implemented enough of it that I needn't worry.

[drammock]

@larsoner LMK if/when you want a more thorough review.