Mark HTTP semantic conventions as stable

[lmolkova]

Fixes #82

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• CHANGELOG.md updated for non-trivial changes.
• schema-next.yaml

[lmolkova]

some limitations we have in the md generator (and can address separately):

• only the original definition (in the registry) has a badge, referenced attributes don't get it
• semconv group in yaml cannot be marked as stable as a whole
• metrics don't have stability levels in yaml

[AlexanderWert]

@lmolkova WDYT about rendering deprecated labels as well? Though I see that deprecated attributes have it also in their brief / description: Deprecated, use ... instead.

[lmolkova]

@lmolkova WDYT about rendering deprecated labels as well? Though I see that deprecated attributes have it also in their brief / description: Deprecated, use ... instead.

They seem to be rendered by default - https://github.com/open-telemetry/build-tools/blob/370e8f443612b312d9d30679f4cd513152fdab1b/semantic-conventions/src/opentelemetry/semconv/main.py#L210

We just never use them in markdown, so they don't appear. If we did, they'd appear as

[trask]

I think this should be "ready to review" as soon as these are done:

• Make url.scheme opt in for HTTP client metrics and make server.port required #459
• Make user_agent.original and http.request.header.* attributes sampling relevant on HTTP server spans #467
• Make client.address sampling relevant on HTTP server spans #469
• Change client.port from recommended to opt-in on HTTP server spans #472
• Change network.protocol.name from opt-in to conditionally required #478

[trask]

@lmolkova can you expand this PR to also update the stability of the http spans page and relevant metrics sections, so that it will be the one PR that we merge related to marking things stable? thx

[Oberon00]

I have been asked to review this PR, I can (only) confirm that as far as I know, there are no blockers left for stability so probably this is good to go. But I haven't been following the spec/semconv work closely enough recently to add a formal review here.