Add cometkim codec - 5,797,365 bytes

[cometkim]

Thanks for the fun golf contest!

Core idea: time-series encoding and out-of-order error correction. This uses events as input, not sorted_events, based on the original data observations.

Perhaps operating without sorting in advance is a bonus? :P

[agavra]

Confirmed in CI! Nice submission, curious why you chose not to sort? there's no time penalty in the golf challenge 😉

┌────────────────────────┬────────────────┬────────────┐
│ Codec                  │           Size │ vs Naive   │
├────────────────────────┼────────────────┼────────────┤
│ Naive                  │    210,727,389 │   baseline │
│ cometkim               │      5,797,365 │     -97.2% │
└────────────────────────┴────────────────┴────────────┘

[cometkim]

I assumed the data was already mostly sorted chronologically. Therefore, I used the data's characteristics as they are. I didn't want to store timestamp values at all.

If we allow comparisons with pre-sorted data, I could remove the errors block, but it wouldn't be able to restore the original event order.

[agavra]

oh i missed that you weren't storing the timestamps at all! I'll have to take a deeper look into your implementations.