Expo-histograms: Zero Threshold configurability.

[MadVikingGod]

What are you trying to achieve?

Is the Metric SDK intended to configure the zero thresholds of Exponential Histograms?

Additional context.

Currently, there are references to the zero thresholds in the Data-Model and in the otlp Protocol , but none in the SDK or API docs.
If this is left unset or set to 0.0, the SDKs would only respect 0.0 or 0 as a zero value, and only external producers could affect this value in the otlp output.

This was raised when implementing the exponential histograms in go: open-telemetry/opentelemetry-go#4245 (comment)

[jack-berg]

Some old conversation about this here.

To summarize roughly, SDKs should default the zero threshold to zero and round all subnormal values to zero:

When zero_threshold is unset or 0, this bucket stores values that cannot be expressed using the standard exponential formula as well as values that have been rounded to zero.

We discussed making it a configurable parameter of the exponential histogram aggregation as being a natural next step, but to my knowledge, nobody has asked for it until this issue.