Neuralynx is over-segmenting data

[h-mayorquin]

This is related to #1773.

We have a file where Neuralynx creates around 20 segments in places where there were None. As in #1770, there was no input to turn the acquisition system on or off; the issue is simply that the gaps are slightly longer than the overly strict requirements of the reader.

I have already prepared an example on gin which I will link in thie issue, where the gaps look like this:

time_s, missing_samples, gap_ms
- 1.637400, 249, 8.300
- 1.650233, 160, 5.333
- 1.659566, 135, 4.500
- 1.668599, 48, 1.600
- 1.674200, 294, 9.800
- 1.688533, 235, 7.833
- 1.782700, 31, 1.033
- 1.787733, 226, 7.533

I know we have the option strict_gap_mode to make the reader more lenient, but in this case it only reduces the number of segments from 9 to 7.

I think we should go further and fully implement the solution proposed in #1773.

[h-mayorquin]

https://gin.g-node.org/NeuralEnsemble/ephy_testing_data/pulls/166