Multi Modal Support

[chr-hertel]

To level up multi modal support, we should support

• Image Input
  • Message Splitting to allow multiple message implementations without stacking arguments #25
• Image Output
  • DallE Support feat: Implement DALL-E image generation for OpenAI Bridge #178
• Audio Input
  • feat: audio input support #189
• Audio Output

With the side goal of managing model capabilities better, currently those LanguageModel::supportsX() methods and Message::hasXContent or MessageBag::containsX() don't scale that well.