(retriever) Add VLM image captioning via vLLM

[edknv]

Description

• Add a .caption() pipeline stage to both batch and in-process ingestors that generates text descriptions for extracted images using a VLM (Nemotron Nano 12B v2 VL via vLLM locally, or a remote NIM endpoint). - Use nv-ingest-api's extract_image_like_objects_from_pdfium_page during PDF extraction to detect, merge, and crop image-like objects (images, shapes, forms) from each page into the images column. - The caption stage filters out small images (< 32px), sends the remaining to the VLM, and writes captions back as images[i]["text"]. Optionally prepends surrounding page text to the VLM prompt via context_text_max_chars.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• If adjusting docker-compose.yaml environment variables have you ensured those are mimicked in the Helm values.yaml file.