Improve native GenAI semantic convention coverage

[eavanvalkenburg]

While adding Agent Framework to the OpenTelemetry GenAI semantic-conventions reference scenarios, we switched the scenario to use native Agent Framework telemetry (no manual reference spans) so the coverage reflects what AF emits today.

Observed native coverage from agent-framework with OpenAIChatClient and sensitive telemetry enabled:

• invoke_agent_internal
• inference
• execute_tool

The native telemetry is already useful, but compared with the GenAI semantic-conventions reference coverage we expected, these pieces are missing or not currently mapped:

1. No GenAI invoke_workflow span

   • Workflow execution emits native workflow spans such as workflow.build, workflow.run, executor.process, edge_group.process, and message.send.
   • These use workflow.*, executor.*, and message attributes, but do not produce gen_ai.operation.name=invoke_workflow with gen_ai.workflow.name, gen_ai.input.messages, and gen_ai.output.messages.

2. No gen_ai.client.inference.operation.details event/log

   • The reference repo has event coverage for inference operation details, but native AF telemetry did not emit this event.

3. Some request attributes appear only on the agent span, not the inference span

   • gen_ai.request.max_tokens, gen_ai.request.temperature, and gen_ai.request.top_p are present on the invoke_agent span.
   • The child chat/inference spans currently include gen_ai.request.choice.count and model/response/usage attributes, but not those request option attributes.

4. Server attributes are partial

   • Native inference currently reports server.address as Unknown against the local mock Responses endpoint and does not include server.port.

This came from running the GenAI semconv reference scenario against AF native telemetry in semantic-conventions-genai, using the local mock OpenAI Responses endpoint and OpenAIChatClient.

[eavanvalkenburg]

Reference coverage PR in OTel semantic conventions: open-telemetry/semantic-conventions-genai#325