Leave the user the option to use Sequence[userContent] in ToolReturn return_value

[all-mute]

Description

Thank you for developing and maintaining pydantic ai!

Context:
When output of (20) tool call are mediafiles, I have situation like that:

All LLM providers recommend to interleave multimodal content.

FR:
Let framework's user ability to use Sequence[userContent] in ToolReturn return_value

References

https://ai.pydantic.dev/api/messages/#pydantic_ai.messages.ToolReturn

[DouweM]

I agree, we're already tracking this in #3253.