Track down why we have the discrepancy between CURIE report and the Redis database

[gaurav]

There is a small discrepancy between the CURIE report and the Redis database: for example, with 2025jan23, the CURIE report has 481,727,857 cliques, but the Redis database has 481,687,260, which means that there are 40,597 more cliques in DuckDB than in Redis. This is almost exactly the number of duplicate CURIEs with have, so that probably explains this (remember that DuckDB allows multiple identical CURIE leaders while Redis does not), but it would be good to know for sure.

We can probably use the DuckDB database to figure out why this is happening.