Python: Feature: telemetry for tracking provider latency

[castlenthesky]

Motivation and Context

1. Why is this change required? - Helps add observability into latency from the LLM provider
2. What problem does it solve? - Current telemetry and observability does not provide any insight into where latency sits during the response generation. This PR aims to add some degree of insight into this question.
3. What scenario does it contribute to? - It contributes to instances where teams are trying to optimize for speed.
4. If it fixes an open issue, please link to the issue here. - Open issue link
   -->

Description

Adding spans/metrics to track streaming latency. Specific metrics added to otel exports:

• gen_ai.client.operation.time_to_first_chunk
• gen_ai.client.operation.time_per_output_chunk

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Pull request overview

This PR adds telemetry metrics for tracking streaming provider latency in the agent framework's observability module. The implementation introduces three new OpenTelemetry metrics to measure streaming operation performance from the client's perspective.

Changes:

• Added three new histogram metrics for streaming latency: gen_ai.client.operation.time_to_first_chunk, gen_ai.client.operation.time_per_output_chunk, and gen_ai.client.operation.duration
• Modified trace_get_streaming_response to track timing information for chunks and record streaming-specific metrics
• Added bucket boundaries configurations optimized for streaming latency measurements
• Created tests to verify streaming metrics are recorded during both successful and error scenarios

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 8 comments.

File	Description
python/packages/core/agent_framework/observability.py	Added new histogram creation functions, bucket boundaries for streaming metrics, modified streaming response tracing to track chunk timing, and implemented _record_streaming_metrics helper function
python/packages/core/tests/core/test_observability.py	Added test fixtures and test cases for streaming metrics recording in both success and error scenarios

[castlenthesky]

@microsoft-github-policy-service agree

[TaoChenOSU]

@castlenthesky Thank you for contributing!

Do you have an ETA on when these two metrics will be merged to the OTel GenAI Semantic conventions?

[castlenthesky]

@TaoChenOSU I don't yet. I just submitted the PR for their repo as well. We can track that status here: open-telemetry/semantic-conventions#3377

Happy to make the updates I flagged with the TODO and submit the final commit once they've officially adopted the convention.

[castlenthesky]

@TaoChenOSU, the telemetry metrics outlined above have been merged into the OTEL standard successfully. I recommend we merge these in as well. What can I do to help move that process forward?

[castlenthesky]

@markwallace-microsoft and @TaoChenOSU Bumping this PR - the changes in OTEL's semantic conventions that support the two telemetry metrics covered in this PR have been accepted and merged into the OTEL standard's main branch. Accordingly this PR should be ready for final review.

[TaoChenOSU]

@castlenthesky Thank you for your contribution!

I left a few comments. Copilot also left a few comments. There are two conflicts. Please address them and I think it's good to go.