Process PowerPoint Slides as a whole image for VLM

[Steve-Allison]

I feel that we've missed the ability to capture what a PowerPoint really is - it's not a bunch of text and graphics on a slide; a slide is a gestalt whole... I would love to be able to run VLM against the whole slide and use that as an image description within Docling, and have the Docling understand that this is what the slide-level image description is.

Currently, I do this as a post-conversion activity and then inject the VLM image description back into the Docling file - but would love to see this as a native capability.