Harmonise sources' existing file checking

[nikbpetrov]

Related to #204

Review what information each source needs to check what to (re)process. #204 sets this up so that we can easily bring in either resolution file ids or full resolution files.

Before #204 , the setup look like so:

For a source, like infer, we load files_in_storage :

 files_in_storage = gcp.storage.list_with_prefix(
      bucket_name=env.QUESTION_BANK_BUCKET, prefix=SOURCE
  )

  dff = source.fetch(dfq=dfq, files_in_storage=files_in_storage)

but those are essentially used as an id filter:

resolved_ids_without_files = [id for id in resolved_ids if f"{self.name}/{id}.jsonl" not in files_in_storage]

Crucially, this operation (because of how list_with_prefix is set up) will list a file even if it's empty.

This is NOT the same behavaviour as the the load_existing_resolution_files function, which explicitly drops empty files from its result:

if not df.empty: `result[question_id] = df

A few options exist:

• listing
• list and check file sizes
• fully loading the actual files