v0.1 — Reproducible RLlib baseline

First tagged release. Focused on making the project's baseline reproducible and its
strongest empirical result easy to find.

What's here

• Quick start: pip install -e . + one command runs a random-policy demo in the
  base environment — no VS Code or Conda required (full VS Code/Conda setup still
  documented for the complete dev environment).
• Headline result: in controlled reward-shaping experiments, a sparse,
  reproduction-only reward beats four denser alternatives on every tested ecological
  outcome — reproduction rate, final population balance, and extinction risk. See the
  README for the chart
  and the full writeup for
  methodology and every module's results.
• EXPERIMENTS.md: the full catalogue of evolutionary and
  fixed-trait environments — Darwinian/Baldwinian evolution, reward shaping,
  cooperation/game-theory environments, and Red Queen coevolution.
• CITATION.cff and CONTRIBUTING.md for citing the project and contributing to it.

Reproduce the headline result

git clone https://github.com/doesburg11/PredPreyGrass.git
cd PredPreyGrass
python -m venv .venv && source .venv/bin/activate
pip install -e .
python ./predpreygrass/non_evolutionary/base_environment/random_policy.py

Then see predpreygrass/non_evolutionary/project_reward_shaping/README.md
for the sparse-vs-dense training configs behind the headline chart.