Reduce the time to generate the OCR field

[liseli]

The following changes were applied to this PR to reduce the time to generate the OCR field.

• Use a list to collect the text, instead of string concatenation
• Remove redundant encoding and decoding operations
• Use content management to open the zip file.
• Remove some redundant logs

In the following picture, you can see how most of the time, the full-text document is in the process of generating the ORC field.

I used seven documents to measure the time to generate the OCR field. Before this PR, around 0.38 seconds to generate the OCR field, after the PR ~0.16 seconds.

Steps to test this PR,

docker build -t document_generator .
docker compose up document_generator -d
docker compose exec document_generator pytest document_generator ht_document ht_queue_service ht_utils

[aelkiss]

Makes sense to me. It looks like before we were copying and appending to the string which would be n^2 in the number of pages, but now we are appending each page data to an array and concatenating once into a string.