Disable reflection after tool use

[WorldInnovationsDepartment]

This PR introduces the reflect_on_tool_call parameter, similar to AutoGen, allowing agents to bypass reflection on tool calls when unnecessary.

Key Changes:
• reflect_on_tool_call Implementation: If set to False, the agent directly returns tool call results without generating an additional LLM response.
• Performance Optimization: Ensures no unnecessary model calls occur when reflect_on_tool_call=False, improving efficiency.
• GraphDeps Refactor: Introduced _build_graph_deps to simplify dependency overrides and make GraphDeps more flexible.

For further details, refer to the related issue: pydantic-ai#1024.

Change inspired to such feature from Autogen framework which I used often and now I am missing in PydAI :(

[aristide1997]

This would be very useful, we still don’t have a way of doing this