feat: multimodal inputs in REPL mode

What is the type of issue?

Documentation missing, maybe

What is the issue?

I see mentioned in the repo that the -i or --image flags can be used such as codex --image "<path_to_image>" to provide an image as input, which works.

However, there doesn't seem to be a documented syntax for adding images in the interactive repl mode, i.e. just codex. I'm not sure if this feature even exists; if so, it could stand to be better documented.

Where did you find it?

I did not see this documented on Github, or in OpenAI's docs here.