feat(ci_visibility): add test_session telemetry metric with provider and auto_injected tags

[gnufede]

Summary

• Adds a new test_session count metric emitted once per test session with provider and auto_injected tags
• Reduces cardinality versus including these tags on the per-test event_finished metric
• Called from the pytest plugin's session finish path alongside record_session_finished

Test plan

• Unit tests for record_test_session with provider and auto_injected tags
• Unit test for unsupported CI (None provider, auto_injected=False)
• Verify metric appears in telemetry intake

Motivation

Per CI Visibility telemetry RFC: a dedicated low-cardinality test_session metric allows tracking CI provider distribution and auto-injection without inflating event_created/event_finished tag cardinality.

🤖 Generated with Claude Code

[cit-pr-commenter-54b7da]

Codeowners resolved as

ddtrace/testing/internal/pytest/plugin.py                               @DataDog/ci-app-libraries
ddtrace/testing/internal/telemetry.py                                   @DataDog/ci-app-libraries
tests/testing/internal/test_telemetry.py                                @DataDog/ci-app-libraries

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4c3d81d767

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Emit test_session metric only from xdist main process

In pytest-xdist runs, pytest_sessionfinish executes in every worker and in the controller process, but record_test_session is called before the if not self.is_xdist_worker guard. This makes the new test_session counter scale with worker count (e.g., 9 emissions for 8 workers) instead of once per logical test session, which skews the provider/auto-injection telemetry this metric is meant to represent.

Useful? React with 👍 / 👎.