Fix/521 pdf2image memory error

[christinestraub]

Summary

Closes issue #521. Implements the same logic as unstructured-inference/PR #136 for the ocr_only strategy.

• Add functionality to convert a PDF in small chunks of pages at a time
• Create a wrapper function convert_pdf_to_images for pdf2image library using Python generator, apply it to _partition_pdf_or_image_with_ocr
• Set the file's current position to the beginning after reading the file in convert_to_bytes

Testing

filename = "example-docs/pdf2image-memory-error-test-400p.pdf"
elements = partition_pdf(filename, strategy="ocr_only")
print("\n\n".join([str(el) for el in elements]))

[qued]

LGTM, I like the generator implementation! One formatting tweak, and a general (nonblocking) suggestion for the function interface.