Revise community ranking algo

[dadofsambonzuki]

The current community ranking 'algo' is weighted on how up-to-date merchant verification is and on the total number of merchants in a given community.

This results in:

1. Big changes on the leaderboard when just 1 merchant becomes out-of-date
2. Large communities by merchant number (i.e. large geographic areas) being ranked higher

I say 'algo' because this logic is split between the star system the API generates based on up-to-date percentiles (20% bands) and the web app, which additionally sorts by number of merchants within a percentile.

I think we should update the current star system with an algo that takes into account up-to-date and community size (either population or area).

This way all 'official' apps will rank communities consistently. Of course other apps are free to implement whatever ranking algo they want based on the core data from the reports.

[dadofsambonzuki]

This algo by @sxajne gives a score between 0 and 1 and takes into account up-to-date and population size. score is a function of community size and so large communities with a single out-of-date merchant are not heavily penalised.

c is 0.5 and if you increase it, large communities (by total merchants) will get less bonus for being large.

I suggest we use something like this to dive a score out of 100 (e.g. 36.76) and give them a 5* rating in 20 percentile intervals to encourage non-linear progression.

[dadofsambonzuki]

I don't think we should include ATMs in the community algo.

Or perhaps weight them in some way.