NeLy-EPFL/flygym ? reverse-engineered prompt

Build me a Python library for simulating a digital twin of an adult fruit fly, like the FlyGym project.

I want a clean package that can run embodied sensorimotor simulations with a biomechanical fly model, visual input, smell input, walking over different terrain, leg adhesion, and basic mechanosensory feedback. It should also support a simple hierarchical control setup where one part handles higher level decisions and another handles motor control, with signals passed between them.

Please make it feel like a real usable research tool, with a nice API, example tutorials, tests, and documentation pages. If there are any recent best practices or current docs for the simulation stack, look them up online if needed.

The main goal is that someone should be able to install it, run a basic fly simulation, inspect sensor inputs and joint state, and plug in their own control logic without fighting the setup.