Top Players to be based on games finished, not games started

[jimmydorry]

• [x] Count number of games abandonned
• [x] Count number of games failed
• [x] Add the usual sorting magic for the display on wins and number of games played (check that we have the right indexes)

https://github.com/GetDotaStats/site/blob/e89e833fb82534757b0d0c453c5671e9c10bfede/site_files/s2/api/s2_phase_2.php#L135-L245

https://github.com/GetDotaStats/site/blob/e89e833fb82534757b0d0c453c5671e9c10bfede/site_files/s2/api/s2_phase_3.php#L184-L240

Do we want to only include stage 3 games for numGamesPlayed? Mods without dedi's will put players at a disadvantage for games played. Including stage 2 games, makes the metric more meaningful, but makes winrate less accurate. Even if stage2 games are included, it would affect everyone equally.

Either way, the metrics are still easily gamed, so accuracy of the results should probably not be the deciding factor.

[jimmydorry]

Base the results summary off of phase 3 only. Still need to make the display of these top players better on the site.