Unified Random Number Generator

[JulianKu]

NEP 19 — Random number generator policy states that regarding numpy.random.* "[the] preferred best practice for getting reproducible pseudorandom numbers is to instantiate a generator object with a seed and pass it around". According to this, the currently used global RandomState in numpy can cause problems. Therefore, this pull request proposes replaces it using the numpy random number generators instead.

With this pull request, a random number generator (rng) object can be passed to the core (with a new default_rng generator object as the default argument). It also adds the same argument to the sample function in cubicasa.py, to the RandomSpawns class and the random_empty_positions function in modules.py, the RandomChain function in fsm.py, and the init_textures, wall_pattern, random_lights, scenery functions in scene.py.

In the RandomSpawns class (modules.py), the core rng is automatically passed to the random_empty_positions during initialization. Likewise, the rng object is passed from scenery to random_lights and init_textures, and from init_textures to wall_pattern.

This way, during environment setup, one can create a single (potentially seeded) rng generator object to pass to all subsequent use cases.