Add more arguments to agent/run hooks

[seratch]

We've been receiving many feature requests to add more arguments to the agent/run hooks. The examples are:

• turn_input to on_agent_start
• arguments to on_tool_start/on_tool_end

and perhaps other data would be needed in the future. However, the current design of the lifecyle hooks does not allow us to quickly add those because changing the method signature could be a breaking change to existing app that rely on the hooks feature.

To get rid of this restriction, we may want to consider accepting both the current set of the arguments and a new unified dict (or similar) object that contains all possible argument data.

see also: openai/openai-agents-js#765