Add statistical aggregation functions

[dominiklohmann]

I needed mean for an ingress/egress chart and was surprised we didn't have that yet. This was easy enough to implement, so I just went ahead and did it.

Based on the initial review feedback this also adds stddev, variance, approximate_median, and collect.

[mavam]

Could you also add something for median and standard deviation, assuming it's as easy? They often go together.

[dominiklohmann]

Could you also add something for median and standard deviation, assuming it's as easy? They often go together.

Bit harder to do, but certainly not a problem. If I find some more time to hack on code this weekend I'll consider it, otherwise feel free to just use mean as a template.

[dominiklohmann]

Could you also add something for median and standard deviation, assuming it's as easy? They often go together.

Done. Mind giving this another review given that there were major changes since Jannis' initial approval?

[mavam]

Great work!

Approving modulo missing integration tests.

I'm a bit worried about the error bounds of using t-digest for incremental median computation, but having even an approximate value is already very helpful. IIRC the error bounds do not hold upon when composing t-digest sketches, but that's not an issue in our case.

[dominiklohmann]

Approving modulo missing integration tests.

I'll add some before merging. For now I mostly verified that this works as expected by manually comparing values. That's why I implemented collect as well.