duplicates can be written to repository visited file.

Welp...

```
/ # wc -l data/visited_repos.csv
321683 data/visited_repos.csv
/ # wc -l repos.csv
223664 repos.csv
```

That's a thing.

Start by locating the Go code that writes data/visited_repos.csv and compare its duplicate handling with repos.csv. Reproduce the count discrepancy, then verify that repeated repository entries are not written more than once and that the visited file remains correct.

Written by the indexing model from the issue text.