Optimize `save_load_state.py` for Performance and Efficiency Improvements

[Mefisto04]

Issue Description:
The current implementation of save_load_state.py in danswer/backend/scripts can be improved for better performance and efficiency. Below are three key areas that need optimization:

1. Postgres Snapshot Handling:

• Current Approach: Uses subprocess to execute shell commands for saving/loading the Postgres snapshot.
• Suggested Change: Use a direct database connection with psycopg2 for more efficient interaction with Postgres, eliminating shell overhead.

2. Vespa Snapshot Handling:

• Current Approach: Fetches documents sequentially with a continuation token.
• Suggested Change: Implement batch processing to fetch multiple documents per request, reducing the number of HTTP requests.

3. Multi-threading for Vespa Operations:

• Current Approach: Sequential processing of documents.
• Suggested Change: Use multi-threading to handle document loading and saving in parallel, speeding up the process.

Please assign me this issue so that I can start working on it.

[Mefisto04]

hey @pablodanswer, please review this and assign me this issue