It's basically an initial set of data populated into a database.

The source code repo contains the database schema itself, so when you stand it up locally, the structure (all the tables) are in place. But, for the most part, they're empty tables. So when you open up SN locally, there's minimal posts, comments, users, etc.

A database seed would populate test data into the database, making it easier to develop or test certain features because you have test/sample data to work with.