fix: strip non-JSON-serializable defaults from @tool schema to prevent PydanticJsonSchemaWarning

[frankgoldfish]

Summary

Fixes #1914.

When a @tool function parameter uses Field(default_factory=...), Pydantic includes the FieldInfo object as the default value in the generated JSON schema. Since FieldInfo is not JSON-serializable, Pydantic emits:

PydanticJsonSchemaWarning: Default value annotation=NoneType required=False default_factory=list description='items' is not JSON serializable; excluding default from JSON schema

Reproduction

from pydantic import Field
from strands import Agent, tool

@tool
def example(
    items: list[str] = Field(default_factory=list, description="items"),
) -> int:
    """Example tool."""
    return len(items)

Agent(tools=[example])  # emits PydanticJsonSchemaWarning

Root Cause

_clean_pydantic_schema() in decorator.py removes known Pydantic metadata keys but does not validate whether default values are JSON-serializable before they reach Pydantic's schema serializer.

Fix

After processing each property in _clean_pydantic_schema(), attempt json.dumps() on any default value. If it raises TypeError or ValueError, remove the default key.

The field remains optional and retains its type and description — only the non-serializable default sentinel is dropped, which is exactly what Pydantic was going to do anyway (with a warning).

Tests

Added test_tool_field_default_factory_no_pydantic_warning regression test that asserts no UserWarning is raised when accessing tool_spec for a tool with Field(default_factory=list).