Revisit package structure in light of multiple model data types

[markpollack]

Currently we have the structure

org.sf.ai.chat
org.sf.ai.prompt
org.sf.ai.prompt.messages

This doesn't feel right for several reasons. First is that message package should not be under prompt as it is more general and used in areas outside prompts, for example the Generation class is of type message.

Message, MessageType and AbstractMessage are very generic in the API, however AssistantMessage, ChatMessage and others are specific to "chat" AI models and should maybe be under the chat package.

Furthermore, we want to support other models that take other types of input and output other than text type. For example, models that support text->image and image->image. Also there is the case of multimodal AI models that can text (text,image,video,audio) -> (text,image,video,audio)

To sort out the best package structure to "grow into over time" a spike should be done, at least with single "modal" model that have a different data type of input/output (e.g text->image, image->image) so see what parts of the API are generic across various model types and which parts are not and should live in package specifc names related to their model type.

[markpollack]

Also some thoughts in issue https://github.com/spring-projects/spring-ai/issues/134 and https://github.com/spring-projects/spring-ai/issues/135

[markpollack]

The PR https://github.com/spring-projects/spring-ai/pull/249 address this.