Investigate parallelizing test discovery

[younata]

Now that the test DSL is a series of methods on AsyncDSLUser and SyncDSLUser, it should be possible to have each class implementing *DSLUser record those tests in the class (For example, instead of having SyncDSLUser.it call into World.sharedWorld.it, it could instead record the tests to some class variable?).

Doing this would allow us to parallelize test discovery on a per-class basis, dramatically improving test discovery runtime.

Before doing this, we should also log to the console how long test discovery takes, so we can get an idea of how long this actually takes and whether it's really worth doing.