Merge the kv and labels packages

[MrAlias]

The kv package contains definitions for keys, values, and their combined key-values with some functionality around their creation and transformations. The label package contains interaction and set functionality around the kv package. Merging these packages into a unified label package would organize related code into a single import and remove the inherent coupling between the packages.

This would also resolve #1010

[MrAlias]

Moving the combined package to the top level directory of the repository would also make this a part of #964.

Additionally, this work should also include updating documentation for the package. The existing package and exported object documentation for both of these packages needs improvement to help new users.

[seh]

One thing that confuses me—perhaps due to ignorance of the OpenTelemetry specification—is that we now have the word "label" applied to "attributes". Is the label what was previously called the attribute key, or the label a key together with a value? If it's the former, are "key" and "label" synonyms? If it's the latter, are "attribute" and "label" synonyms?

[MrAlias]

One thing that confuses me—perhaps due to ignorance of the OpenTelemetry specification—is that we now have the word "label" applied to "attributes". Is the label what was previously called the attribute key, or the label a key together with a value? If it's the former, are "key" and "label" synonyms? If it's the latter, are "attribute" and "label" synonyms?

It's the latter (a label is a key-value pair), "attribute" and "label" are synonymous. Attributes are used in the tracing specification (though have recently been moved to a "common" part of the specification) and labels are used in the metric specification.

I agree this is not ideal.

[seh]

Thank you for the clarification. Are you aware of any motion (even rejected) to unify those terms? I recognize that would yield yet more churn in the code. The current state comes after several big moves, but it's still muddled.

[MrAlias]

Thank you for the clarification. Are you aware of any motion (even rejected) to unify those terms? I recognize that would yield yet more churn in the code. The current state comes after several big moves, but it's still muddled.

There was an old conversation in the specification SIG around this issue that was left with the different names being left as is (IIRC this was because no resolution could be reached and it was the default).

However, in this repository I've wondered if we can do better. I have a lot of thoughts on possible approaches, but I think they are better suited for a new issue: #1069