Support for native histograms

[jml]

Native histograms (aka sparse histograms, aka high-res histograms) are on their way to being production ready. It would be wonderful if the Node prom-client supported them.

From the Prometheus docs:

A native histogram requires only one time series, which includes a dynamic number of buckets in addition to the sum and count of observations. Native histograms allow much higher resolution at a fraction of the cost.

They aren't quite a stable feature yet, but I figure it would be useful to me & others to have an issue to gather together any discussion on implementing them for this client.

Some relevant links:

• Prometheus milestone for full native histogram support
• PromCon 2022 talk introducing native histograms
• Go client implementation of histograms, including support for native histograms

[equinsuocha]

I think it worth discussing again, as native histograms are considered a stable feature, so why not implement it?
Some ground work on protobuf metric exposure has been done quite a while ago in addition to the native histogram spec and reference implementations.

[nicolas-fidel-wmx]

Hi @jdmarshall
With the changeover to the official Prometheus team, do you have any news regarding the integration of native histograms?