Update backfill scripts for classifications

[yuenmichelle1]

Keeping both old script and new script for record.

Old script (where we use straight up COPY from source and COPY TO target db in one go), ended up taking a long time but also query cancels out and connection to db/s is lost. Because of how COPY works, its hard to debug where connection gets lost and because its an all in situation, we get no data copied. [this is backfill_classifications.py]

To mitigate, we did the following:

• split up the connections so that:
  1. from the source db, we save it all to csvs containing about 10,000,000 rows.
  2. once all csvs are saved, close connection from source db, since we have everything we need from it.
  3. connect to target db and copy from all the csvs saved.
  4. then close connection from target db
     [This is in backfill_classifications_chunk_in_files.py save_classifications_chunk_in_files.py then copy_classifications_from_files.py ].

Some minor changes:

• in backfill_classifications.py, moved the long 🍑 query into its own line.
• updated the query to take care of the case where a classification's metadata.user_groups has null in it. (eg. [1234, null, null] ). This has shown up in some classifications while testing.

[zwolf]

Looks good, though if it were me I'd probably be less inclined to run the export and the import as a single script. Since you're creating the files anyway, you have the opportunity to create them all and confirm success, then be done with that step. Your call of course but in light of previous mystery issues I'd probably break this into two scripts/steps, one to create the files and one to read them.