The old test runner (currently used in CI) had a regression when moving

to pytest. It failed to convert the first argument (the database name to

run everything against) to the new pytest --db= argument.
This change adds this back to the test runner. If an argument is passed,

and isn't --, it's expected to be the database name as before. All

other arguments are then passed as normal.

Ran the old test runner with/without a database name, with/without --,

and with/without specific tests to run.