PR Description

This code fixes handling of optional arrays in pydantic models provided as tool inputs like this:

class ExampleToolInput(BaseModel):
        numbers: Optional[List[str]] = Field()

Optional parameters are handled by setting nullable field in gemini API. Current code works for every type, except for arrays, because items field, that contains information about type of values inside the array, is present in anyOf field as one of the possibilities, where it's ignored.

Also added log where user is warned that only only type can be used inside array. Before, last defined type was used.
This is an edge case where user could be trying to use an array with multiple types which is not supported by genai, but can be useful for finding bugs

Type

🐛 Bug Fix
✅ Test

Testing(optional)

Launching pytest with arguments test_function_utils.py::test_tool_input_can_have_optional_arrays --no-header --no-summary -q in /xxx/langchain-google/libs/genai/tests/unit_tests

============================= test session starts ==============================
collecting ... collected 1 item

test_function_utils.py::test_tool_input_can_have_optional_arrays 

========================= 1 passed, 1 warning in 0.61s =========================