Multimodal support in another ChatGenerator

[anakin87]

A suggestion would be Anthropic or Bedrock.

This would mean converting ChatMessage with ImageContent parts to the specific format
of the chosen LLM.

Depends on #9258.

[sjrl]

We have a draft implementation for Bedrock here. Even though this is for Bedrock, the draft only implements support for Anthropic vision via an AnthropicClaudeVisionAdapter

Also it should be basically free to get vision in AzureOpenAIChatGenerator once it's finished for OpenAIChatGenerator.