humanized leaderboard numbers should probably take into account all the data

[hamilton]

I think it's a good idea for us to figure out how to avoid cases like 1345.4k (would expect 1.34m) in the leaderboards and 40.0k in the axes (would expect 40k)

the existing heuristic may need some tweaking. Could we check what the legacy dash does?

[djbarnwal]

@hamilton Tried taking all data into account. Even though the scale chose by the algorithm remains to be k. This is happening because the nulls are now skewing the results a bit. Do you think adding commas would make it look better to instead of 1353k we can show 1,353k?