fix: align parser span kind with OpenInference

[radu-mocanu]

Summary

• ensure parser spans use the OpenInference CHAIN classification consistently
• keep OpenInference within the 0.1.x compatibility range
• apply the two-day package quarantine consistently across integration testcases

Why

OpenInference 0.1.69 changed parser spans from UNKNOWN to CHAIN in Arize-ai/openinference#3486. Requiring that version keeps runtime traces and test expectations deterministic.

[Copilot]

Pull request overview

This PR updates the project’s OpenInference LangChain instrumentation dependency to ensure parser spans are classified as CHAIN (per the OpenInference contract change), and aligns the ticket-classification trace expectations accordingly.

Changes:

• Bump openinference-instrumentation-langchain to >=0.1.69 (and lock to 0.1.69) to match updated span-kind behavior.
• Update the expected trace fixture to assert PydanticOutputParser spans are openinference.span.kind = CHAIN.
• Increment package version to 0.15.4 and keep uv “exclude-newer” settings aligned.

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated no comments.

File	Description
pyproject.toml	Raises the minimum OpenInference LangChain instrumentation version and bumps the package version to 0.15.4.
uv.lock	Locks the updated dependency version and reflects the project version bump in the editable package entry.
testcases/ticket-classification/expected_traces.json	Updates span-kind expectation from UNKNOWN to CHAIN for parser spans to match OpenInference behavior.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud