[FEATURE] Support Multi-Modal input to __call__

[zastrowm]

Problem Statement

I would like strands to support passing in more datatypes to Agent __call__ so that the images, audio, etc can be passed into the agent

Proposed Solution

This should support a ToolResult as input

agent = Agent()
agent(ToolResult(...))

This should also allow for an empty input to invoke, and it will invoke the current messages array.

agent = Agent(messages=[...])
agent()

Use Case

Passing in images or text or both would allow responding to agents needing more than just text

Alternatives Solutions

No response

Additional Context

No response

[theagenticguy]

#60 adds a structured_output method to Agent, leaving a note that we should consider multi-modal input for that case as well. For example, if I send an image and I want to extract certain image attributes according to a Pydantic JSON schema.