feat: extract parameter descriptions from function docstrings

[heath-hunnicutt-ruach-tov]

Summary

• Parse Google, NumPy, and Sphinx-style docstrings to automatically populate parameter descriptions in tool/prompt/resource JSON schemas
• Previously, parameter descriptions required explicit Field(description=...) annotations; now they can come from standard docstrings
• Field(description=...) annotations take precedence when both are present, ensuring full backward compatibility

Closes #226

Changes

• New: src/mcp/server/mcpserver/utilities/docstring_utils.py — docstring parsing utility using griffe with auto-detection of Google/NumPy/Sphinx formats
• Modified: func_metadata() accepts optional param_descriptions dict, injected as fallback when no Field(description=...) exists
• Modified: Tool.from_function(), Prompt.from_function(), ResourceTemplate.from_function() — parse docstrings and pass results through
• Modified: pyproject.toml — added griffe>=1.0 as runtime dependency
• New: 8 unit tests for parsing utility + 7 integration tests covering all formats, precedence rules, and edge cases

Test plan

• All 8 test_docstring_utils.py tests pass
• All 8 test_parameter_descriptions.py tests pass (1 existing + 7 new)
• All 243 existing mcpserver tests pass with zero regressions
• CI passes on all supported Python versions (3.10–3.14)

🤖 Generated with Claude Code

[maxisbey]

Thanks for the PR! We're going to hold off on this — Pydantic is extracting their docstring parsing logic into a separate package (see Kludex's comment on #226) and we'll use that instead of adding griffe as a dependency.

_{AI Disclaimer}