Add LLaVA OneVision model support

[RyanJDick]

Summary

This PR adds support for the LLaVA OneVision model type:

• The recommended model is available under the "Starter Models" list.
• The LLaVA OneVision VLLM invocation can be used for inference. It supports 0-3 input images along with an input prompt.

Example

Output:

The image is a digital illustration that depicts a surreal landscape with a prominent water tower in the foreground. The tower is tall and cylindrical, with a platform at the top that has a railing. It is surrounded by a grassy field with small white flowers. The sky is filled with various celestial bodies, including a large moon and several smaller moons, creating a dreamlike atmosphere. The clouds are fluffy and scattered across the sky, and the overall color palette is warm, with shades of orange, pink, and blue dominating the scene. The art style is reminiscent of a science fiction or fantasy genre, with a focus on imaginative and fantastical elements.

Related Issues / Discussions

N/A

Remaining Work

• Add the new model type to the frontend so that it appears in the Models tab.
• Add a model identifier input to the LLaVA OneVision VLLM. Or, only support a single model and raise if it's not installed with a reference to the starter model.

QA Instructions

• Test model installation via starter model list
• Test that installed LLaVA models appear in the model list.
• Test inference with 0 images
• Test inference with 1 image
• Test inference with 2 images

Merge Plan

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Tests added / updated (if applicable)
• Documentation added / updated (if applicable)
• Updated What's New copy (if doing a release after this PR)