Patch to fix uconn timestamp storage

[lisaSW]

This is a patch to store every timestamp in the uconn ts list rather than a unique set. This will not impact any of the scoring algorithms, which already gather the timestamps with $addToSet

[Zalgo2462]

The code looks good. I will give it a run here in a bit.

We might want to consider using TsListFull instead of TsList when calling getTsHistogramScore on line 218. This would keep the timestamp list in line with the timeline/ histogram scores.

It would be convenient if we could also use TsListFull for the rest of the scores. Below, I've written a bit about using TsListFull for the ts skew and dispersion scores. If you see any reasons which would prevent us from doing so that I have not addressed below, please let me know.

From a first glance, it looks like the Bowley skew would actually work with tsListFull. If we use TsListFull, then tsLow, tsMid, and tsHigh might all be 0. If just tsLow is 0, then it doesn't look like we will run into any issues. In the case that both tsLow and tsMid are 0, we have a special case that makes sure that tsSkew gets set to 0 on line 137. The skew score is defined as 1 - tsSkew, so a beacon with all of its delta times within half a second would score high on skew. I think that is a reasonable expectation.

It looks like the issue with the dispersion score hits when we go to create the tsMadmScore on line 183. We divide by tsMid here which could be 0 if we use tsListFull. I believe the only distributions where we'd see tsMid get set to 0 would be distributions with a huge spike at 0 and a tail of smaller bins out to the right. In these cases, the median deviation from the median should also be near 0. So, I think there is good reason to add a special case here where we just set the tsMadmScore to 1 if tsMid is 0.

[Zalgo2462]

We discussed using TsListFull for IP beacon scoring, but that can be implemented in a separate PR if the testing goes well. I tested this PR as it stands and it works well.