Add optional template variability datasets (std, median) to the Templates spec?

[h-mayorquin]

SpikeInterface's ComputeTemplates extension computes variability arrays (std by default, optionally median, mad) alongside the average templates, but the current NWB write step only stores the average and discards the rest. Would it makes sense to add those as optinoal datasets? Does the GUI uses?

@alejoe91