Text to Image: Additional operations

We should look at plumbing additional common operations and how those will work.

We've started out with text + image? + options. This handles generate, edit, variations, upscale.
- generate: text
- edit: text+image
- upscale: text=null + image + options.size != null
- variations: text=null + image + options.size == null

We don't yet have a happy path for the following operations which are somewhat common:
- Inpaint
- Outpaint
- Erase
- Find / Replace
- Multiple input images

Most can likely be achieved through RawRepresentationFactory but we shouldn't do that if they are common enough to have a first-class cross-provider experience.

References:
https://platform.openai.com/docs/api-reference/images
https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html
https://platform.stability.ai/docs/api-reference