Decouple player rank from percentiles

[Stuart1998]

Having player rank be tied to percentile sounds great in theory but in practice it causes problems.

1. It leads to rank thresholds fluctuating wildly based on how long it's been since a server restart. This is because players are only ever removed from the ladder after a server restart and these players are severely disproportionately those who were previously lower on the ladder. As a result, a player who was 20th on the casual ladder before the restart is probably still 20th after the restart, but 20th out of ~1600 instead of 20th ~3000. This is frustrating for players who were close to ranking up before the restart as all of their progress feels like it's been wiped away.
2. Casual ladder is perpetually far more populated than competitive ladder and skewed towards lower skilled players compared to competitive ladder. I believe TAW being less intimidating than matchmaking/pro 1v1 host and private pvp games always being casual rating (if they're not noelo) are both major contributors to this, but regardless the result is that the rating thresholds to reach a certain rank are dramatically higher for competitive ladder than casual ladder. To reach neutron star (blue) rank in casual ladder, a player needs to have a rating of ~2420 (without accounting for promotion thresholds). To attain it in competitive ladder, a player needs a rank of 3k, and this is without even accounting for somewhat increased WHR inflation in casual ladder than competitive ladder (see Godde having ~400 rating higher in casual ladder right now than in competitive). Note that this has been exacerbated by a server reset in the past week (see Server reset impact dramatically higher in 1v1 ladder than casual ladder #3049) but even without that skewing the numbers even harder it's still dramatically easier to attain a given rank in casual ladder than it is in competitive ladder.
3. Percentile based demotion thresholds result in players in the lower ranks never demoting. With 1812 players in the casual ladder right now, you'd expect the bottom 362 to be in the lowest rank (nebulous). The top 100 of those ranks (1451-1550) are fairly evenly split between nebulous and higher ranks; over the next 232 players, 225 of them are nebulous rank, as you'd expect. However, a majority of the last 30 players, in other words, half the 30 worst active Zero-K players in the world (bottom 1.7%) aren't in the bottom rank; 10 of them are in Brown Dwarf (nominally top 80%) and 6 of them are in Red Dwarf (nominally top 60%). Some of these players seem to be getting their rank from competitive ladder despite most of them being inactive on it, but others have never played a single game that counted for another ladder (see davarit and HuggeTrugge). Rather, these players formerly qualified for a higher rank then dropped dramatically in rating, but haven't deranked because their ladder rating is still within the 20% threshold, due to the tight rating differences at the lower end of the ladder.

To solve these problems, rank should be switched to new numbers based entirely on raw ladder rating regardless of playercounts. A couple months ago I proposed the following numbers as supposed promotion/demotion thresholds for each rank. I think they hit the ballpark well but I don't like how arbitrary they are and would be open to something more formulaic.

Singularity: 3050 promotion, 2950 demotion
Neutron Star: 2450 promotion, 2350 demotion
Supergiant: 2050 promotion, 1950 demotion
Giant: 1850 promotion, 1750 demotion
Subgiant: 1650 promotion, 1450 demotion
Red dwarf: 1400 promotion, 1200 demotion
Brown dwarf: 1200 promotion, 1100 demotion