RFC: Automatically apply a time-based decay to the access statistics

Our long-running top dataset with the most accesses which appears relatively random is probably a good example of an echo chamber effect created by popularity-based ranking of search results.

A simple yet hopefully effective means to mitigate this issue might be a time-based decay of the access counters. The main assumption being that genuinely popular datasets will have their accesses outgrow the decay. Additional, the slow rate of decay still allows for transiently popular datasets.

This could also be beneficial from a technical point of view as it keeps the access statistics relatively sparse and therefore easier to store and manipulate.