WIP: Use greater distance when defining overlapping nirs channels

[rob-luke]

What does this implement/fix?

When you have a NIRS recording with lots of very close channels the topomap looks bad. We had code to handle overlapping NIRS channels, but it was for exactly overlapping channels. This extends what's considered overlapping channels to 1.5cm. In NIRS experiments the source and detectors are ideally placed 3cm apart, so I have taken half this value and it looks nice in my figures.

Example of topoplot with existing value (note the small black dots representing channels in little clusters)...

And with the new value

Additional information

1.5 cm seems to play nicely with the topographic plotting too. But Im open to other suggested values or perhaps a totally different approach?

[rob-luke]

The tests for this will fail due to the bug in #8306. I will rebase once that's addressed.

[larsoner]

I wonder if a more general approach would be to allow smoothing the topomaps. Can you see if tweaking these lines to use Rbf interpolation with some suitable smooth > 0? Basically I think we could add smooth=0 as a default EDIT: which would continue to use CloughTocher, and values > 0 switch to using Rbf with the given smooth value.

[larsoner]

https://github.com/mne-tools/mne-python/blob/master/mne/viz/topomap.py#L609-L615

[rob-luke]

Great, I will play around and report back.