Enhancements to update stream conversion script

[szarnyasg]

A followup of #203:

• Old raw data sets have hasInterest instead of TagId for the Person_hasInterest_Tag table. This is the only difference between the data sets, so re-generating everything is hardly worth it... Rather, the script could/should auto-detect this.
• The batching should be done on a daily basis (opposed to a weekly one), otherwise generating the updates for SF10,000 would require 1.5TiB memory.

[GLaDAP]

Would it be feasible to change the column name of the old raw datasets possible using something like:

echo "COPY(SELECT creationDate, deletionDate, PersonId, hasInterest as TagId FROM '*.parquet') TO 'Person_hasInterest_Tag.parquet' (FORMAT 'PARQUET')" | duckdb

[szarnyasg]

They are compressed in tar.zst then split in .tar.zst.000, so, yes, but at that point I'd rather just regenerate the whole thing on EMR.