Add several tests, with related fixes.

This MR adds more tests (in fact it adds tests for all commands, but many of them are still not implemented, and therefore skipped). Also corresponding fixes in the db device. We now have tests for around 50% of commands, and most of the "common" are covered as far as I can tell.

It also adds a few features to the test runner, and documents it in the readme.

As the test runner becomes more complex, the value of describing the tests as JSON shrinks. I still kind of like it, as it's very structured and readable. Also, if we need to do more complex tests we can always write them in python instead, while using the JSON format for the "simple" tests. But it can be discussed.