Add support for using existing page images for VLM file parsing

[zmccormick7]

Useful if you want to create the page images first (still using the FileSystem class) so you can do other sorts of VLM processing with them, such as document type classification, before using them for RAG.