Re-add source_ prefix to archived source files

[mikeknep]

In the PR moving source data from memory to disk, I introduced a bug in the creation of the source data archive (source_tables.tar.gz) that broke the restore-from-backup functionality.

Original backup-and-restore behavior

• Write each source table from Pandas DataFrame to a local CSV named source_{table}.csv
• Archive those files, upload the tar to the project as a project artifact
• When restoring from backup file, pull down the archive, unpack, and read those files back into DataFrames

Bug from previous PR

Since the tables now exist on disk and we only store pointers to them in memory, we don't need to call dataframe.to_csv; instead we just add the existing local files to the archive. The problem is that the filenames of the archived files changed—they no longer had the source_ prefix but were instead just {table}.csv:

Consequently, when restore pulls down and unpacks the archive, it looks for files that do not exist (e.g. it tries to find source_users.csv but the file in the archive is just users.csv).

The fix (this PR)

Add the source_ prefix to each table filename when adding it to the archive.

[gracecvking]

Easiest code change to review 👍