Auto-prune BM database to enforce age-dependent snapshot density

To reduce BM database growth, the list of snapshots in the database could be pruned regularly to enforce an age-dependent density profile: The older the snapshot, the lower the local density (density = average number of snapshots reported per day).

Note that a high snapshot density is likely to be more useful for new results than for old ones: The higher the snapshot density, the less effort is required to locate the exact revision that introduces a regression (using 'git bisect' etc.). On the other hand, the longer a regression has gone unnoticed, the less likely you are to care about it anyway (i.e. important regressions should normally be fixed soon after they happen).

One problematic issue with this technique is that it would affect change scores (in particular the durability score which is affected by the length of a contiguous range of significantly equal observations). However, the more stable a change is, the less likely it is to be affected.