Draft: RFC: Extend access stats to collect dwell time CDF

This uses a bit of JavaScript and two new routes to include an approximate CDF for the dwell time (>=10s, >=1min, >=5min) on the dataset pages as show case for how definining and implementing such privacy-respecting stats would look like.

I would not want to merge this as I don't think we need these measurements for the dataset details pages, but doing it this meant it took 30 minutes to implement as it piggy-backs on the existing stats storage we already use to collect access counts.

Furthermore, whether one uses this coarse-grain pseudo-CDF or a more detailed dynamic set of histograms buckets is immaterial as far as I can see, but I actually do like that such an approach makes it easy to define relatively targeted statistics instead of collecting "raw" data.