Fix tool-call fallback for structured completion content

[suger-m]

Summary

• Fix structured response parsing when provider responses return empty assistant content and place JSON in function tool-call arguments.
• Add fallback in LLMCompletionResponse.content to read tool_calls[].function.arguments when message content is empty.
• Update gather_completion_response and gather_completion_response_async to use response.content.
• Add unit regression tests for content fallback, gather helpers, and structured parsing.

Issue

Fixes #2263

Validation

• uvx ruff check packages/graphrag-llm/graphrag_llm/types/types.py packages/graphrag-llm/graphrag_llm/utils/gather_completion_response.py tests/unit/utils/test_completion_response.py
• python -m pytest tests/unit/utils/test_completion_response.py -q

[suger-m]

@microsoft-github-policy-service agree

[microsoft-github-policy-service]

@suger-m the command you issued was incorrect. Please try again.

Examples are:

@microsoft-github-policy-service agree

and

@microsoft-github-policy-service agree company="your company"

[suger-m]

@microsoft-github-policy-service agree company="your company"