Python has bindings for pretty much everything, so I'd imagine that the database you're most familiar with, or that there is the most information on (most likely MySQL, Postgres, or SQLite)
I'd suggest SQLite unless you had a ton of users interacting all at once (if it's something involving editing excel spreadsheets, I doubt you'd have a problem with this).
Slightly off topic, but has anyone built an Excel-esque program on top of SQL? I feel like something of the sort could really be an improvement, as now it'd be easy for end users (who normally use excel) and developers to collaborate on the same data.
Like the article says, SQLite is a good choice when you're starting out. It's included in Python and is great for the basics.
Once you have a lot of data, it really depends on your specific situation. If you don't have any preferences, you can try something new - MongoDB is interesting.
MongoDB/NoSQL is not for the type of analysis mentioned in the article and is NOT a replacement for a spreadsheet.
Out of all the nosql options, why pick the one with the most broken default configuration? People complain that MongoDB is has a bad rep when used for things its not meant to, and then its constantly offered as solution.
And works with Python.