Implement IngestedImages in LLM Engines

[ngupta10]

1. Either there should be a relevant entity in the figure itself
2. Or it has a minimum of 2 entities in the same sentence on the page text.

[saraswatpuneet]

I think we want

• If we find one entity in ocr text extracted from image OR
• if atleast two entities are identified on same page

Check for OCR first and if result is false then second but not the other way round to save processing time