Add multimodal support for the new ChatMessage class

[LastRemote]

This is a to-do for myself to add multimodal supports (image, audio and pdf for OpenAI and Anthropic models) in the experimental implementation of ChatMessage according to deepset-ai/haystack#7848 (comment)

Will open a draft PR later this week.

Tasks

Beta Give feedback

1. Update ByteStream to support automatically loading mime_type
2. Add a new MediaContent to ChatMessageContentT
3. Serialization & Deserialization for ChatMessage and ByteStream
4. Update message_to_openai_format
5. Update message_to_anthropic_format
6. Create a draft PR