Handle external signals to topic-extract gracefully

[jankrepl]

🚀 Feature

It would be very useful to be able to send a signal (e.g. SIGINT / keyboard interrupt) to a running bbs_database topic-extract process and have a guarantee that the already processed articles would be stored inside of the output .jsonl before the process is killed.

Currently, one needs to wait for all articles to be finished.

Why useful?

• Debugging purposes - e.g. timeout 100 bbs_database topic_extract .... I am currently working on the overall pipeline and it would be nice to be able to have "some results" quickly rather than having "all results" slowly.

Other commands at least have a logic of saving things to disk when they run (download, parse) so killing them is not a big deal.

[jankrepl]

A couple of related ideas:

• Write to the .jsonl more regularly (maybe not necessary)
• Check whether any entries already exist in the .jsonl and if yes, make sure to skip those articles