Add multimodal example

[mscheong01]

Feature Description

Add example for multimodal capabilities

Motivation

#5882 took out the multimodal features from the server. Given it's a highly requested feature, our plan would be to reintroduce it at some point (#6168). How about we set up a solid multimodal example elsewhere and then port it to the server example later on?

Possible Implementation

Implementation based on the removed code from https://github.com/ggerganov/llama.cpp/pull/5882/files which had already implemented this feature in the server.cpp example, hopefully with some performance optimization.
For the example, image file could be provided via command line option.

[phymbert]

Have you read:

• server : improvements and maintenance #4216

This looks duplicate of:

• llava-cli: improve llava-cli and the API for using LLaVA #6027

[mscheong01]

@phymbert Ah, I see, I guess llava-cli fits my description for a solid multimodal example. Thanks for pointing it out 🙇‍♂️