feat: reuse documents/embeddings for files with the same sha256sum (+fix pgx bug)

[iwilltry42]

• calculate sha256sum for the content of each ingested file and record it in the DB
• when ingesting a file, check if there's a file with the same checksum in the DB
• if yes and the embeddingModel (config on the owning dataset) matches the incoming one, then fetch the file's documents and their embeddings to re-use for the incoming file

NOTE: This is fine in the Obot setup, as we're using a static flow config for Obot, but if someone is playing with the flow config to test different settings, this may lead to reusing documents from file that were ingested with an old config - we can add checks for that as a follow-up.

Also, I figured that there's a race condition in the pgvector implementation, because appending to the pgx batch queue is not lock protected and I saw some cases, where documents got lost due to this. This is fixed now.

Issue: obot-platform/obot#1803