Better unit test

[ingambe]

We need to coordinate on better unit tests.

The first test is a random one, where we simply test random legal action and hope for no crash.
It also helps check state limits (check that the env state output value between 0 and 1) and test the logic.
TODO: It could be cool to call Clingo on the random sol and check correctness.

Another type of test check that literature lower bound works within our environment.