Document using toolsets directly with PydanticAI in any task

[kaxil]

• Add a "Using Toolsets Directly with PydanticAI" section to toolsets.rst documenting that toolsets work anywhere you can run Python within Airflow (@task, PythonOperator, custom operator execute()), not just via AgentOperator or @task.agent.
• Add an example DAG showing @task + PydanticAIHook.create_agent() + SQLToolset.
• Explain the tradeoffs: full agent lifecycle control vs losing durable execution, HITL review, and automatic tool logging.

@task
def analyze():
    import pydantic_ai

    agent = pydantic_ai.Agent(
        "anthropic:claude-sonnet-4-6",
        system_prompt=(
            "You are a sales analytics assistant. "
            "Use the available SQL tools to query the database and answer questions. "
            "Always start by listing tables and checking their schema before writing queries."
        ),
        toolsets=[SQLToolset(db_conn_id="sqlite_default")]
    )
    result = agent.run_sync(
        user_prompt=(
            "Which customers have spent the most in total? "
            "Show the top 5 with their total spend and number of orders. "
            "Also tell me which product category generates the most revenue."
        )
    )
    return result

[eladkal]

@kaxil sent you a question in Slack