Commented out code for debugging fitting collisions

[JoeZiminski]

Small thing but there is is some commented out code for debugging the spike collisions here. It would be nicer if this code was uncommented and made non-public with a leading underscore. Then their purpose for debugging can be explained in the docstring.

EDIT:
It would also be nice to extend the docstring of the related classes ComputeAmplitudeScalings and AmplitudeScalingNode. I'm having a hard time understanding why this this amplitude scaling is performed (please let me know if I have missed anything in the docs).

[alejoe91]

Hey @JoeZiminski

I can extend the docs and uncomment the functions in fixing #2882

[JoeZiminski]

Brilliant thanks @alejoe91 I was looking at #2882 that made me think about this, let me know if I can help with anything!

[alejoe91]

Well if you have time you can have a stab! My plan was just to protect the LinearRegression function and set scalings to NaN for these spikes.

Regarding what this function does:
The idea is fitting templates to each spike and return the amplitude scaling. This is better than spike_amplitudes, which simply take the voltage value at the spike time, since the latter fails when there is a collision. In addition, also without a collision, fitting the template to the traces snippet is a more robust estimate of amplitude.

[JoeZiminski]

Sure I'll have a go!

[JoeZiminski]

Adressed in #2893