feat: Align retrieval spans with OpenTelemetry semantic conventions

[Cirilla-zmh]

Description

Summary

Align retrieval spans with the LoongSuite Retriever semantic convention and introduce a typed RetrievalDocument model.

Changes

Semantic Convention Updates

• gen_ai.operation.name: retrieve_documents → retrieval
• gen_ai.retrieval.query → gen_ai.retrieval.query.text for query text
• Span name: retrieval {gen_ai.data_source.id} when data_source_id is set
• Add RetrieveInvocation fields: data_source_id, provider, request_model, top_k

New RetrievalDocument Dataclass

• Typed representation with id, score, content, metadata
• RetrieveInvocation.documents accepts List[RetrievalDocument] only
• Instrumentation converts framework types (e.g. LangChain Document) to RetrievalDocument

Serialization (ToolDefinition-style)

• When content capturing is NO_CONTENT: only id and score are serialized
• When content capturing is SPAN_ONLY or SPAN_AND_EVENT: full document is serialized

LangChain Instrumentation

• Add _documents_to_retrieval_documents() to convert LangChain Documents to RetrievalDocument
• Update retriever spans to use the new attributes and span naming

Testing

• Updated util-genai and langchain retriever tests
• 14 files changed, 375 insertions, 154 deletions

Fixes #144 (issue)

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Add unit tests

Does This PR Require a Core Repo Change?

• No.

Checklist:

See contributing.md for styleguide, changelog guidelines, and more.

• Followed the style guidelines of this project
• Changelogs have been updated
• Unit tests have been added
• Documentation has been updated