core: make docstring parsing more robust

[ccurme]

Privileged issue

• I am a LangChain maintainer, or was asked directly by a LangChain maintainer to create an issue here.

Issue Content

Valid Google-style docstrings with descriptions > 1 paragraph are not supported.

Example failure:

from langchain_core.tools import tool


@tool(parse_docstring=True)
def search(query: str) -> None:
  """Searches for information based on the query.

  Use this tool to look up factual information or search for relevant content.

  Args:
    query: The search query. A good query is a clear and concise question or statement.

  Returns:
    A list of documents that match the query.
  """
  pass

Root cause is raising an error when second \n\n-delimited block does not start with Args:

langchain/libs/core/langchain_core/utils/function_calling.py

Line 618 in ec9b414

len(docstring_blocks) < 2 or not docstring_blocks[1].startswith("Args:")

[isatyamks]

@ccurme Could you please review my PR #28730? I improved _parse_google_docstring to support multi-paragraph descriptions while ensuring backward compatibility.