Delete severely undercounted points, caused by script/runtime error

As recently discovered and debugged, the active-user-count script did not always run and finish as designed. So, the counted numbers of users were more or less underestimated in the recent months.

I propose to delete obviously undercounted months, to get a cleaner curve for presentations etc.

👉 !33 (diffs)

I vote for deletion of the most severely affected months, since the numbers a) were never overestimated, only underestimated; and b) better to have no data than wrong data.

---

Current plot:

With severe dips on Sept 1 2023, Oct 1 2023, March 1 2023, and April 1 2023