feat: add OECS.PlayTest project

Introduce a new PlayTest project containing tools for simulating
gameplay, including:

- Agent abstractions (IAgent, GreedyAgent, WeightedAgentPool)
- ObservableCapture for logging entity changes
- PlayLog for generating and saving formatted play logs

OECS.PlayTest/Agents/GreedyAgent.cs

@@ -0,0 +1,54 @@
+using OECS;
+
+namespace OECS.PlayTest;
+
+/// <summary>
+/// Abstract agent that picks the action with the highest score.
+/// When <see cref="ScoreAction"/> returns 0 for all actions (the default),
+/// the agent behaves as a uniform random agent over legal actions.
+/// Ties are broken randomly.
+/// </summary>
+public abstract class GreedyAgent<TDecision> : IAgent<TDecision>
+{
+    private static readonly Random _rng = new();
+
+    /// <summary>
+    /// Returns the list of currently legal actions.
+    /// </summary>
+    protected abstract List<TDecision> GetLegalActions(World world);
+
+    /// <summary>
+    /// Scores an action. Defaults to 0 (uniform random).
+    /// </summary>
+    protected virtual float ScoreAction(World world, TDecision action) => 0f;
+
+    public TDecision Decide(World world)
+    {
+        var actions = GetLegalActions(world);
+        if (actions.Count == 0)
+            throw new InvalidOperationException(
+                $"{GetType().Name}: no legal actions available");
+
+        if (actions.Count == 1)
+            return actions[0];
+
+        float bestScore = float.MinValue;
+        var bestActions = new List<TDecision>();
+        foreach (var action in actions)
+        {
+            float score = ScoreAction(world, action);
+            if (score > bestScore)
+            {
+                bestScore = score;
+                bestActions.Clear();
+                bestActions.Add(action);
+            }
+            else if (score == bestScore)
+            {
+                bestActions.Add(action);
+            }
+        }
+
+        return bestActions[_rng.Next(bestActions.Count)];
+    }
+}

OECS.PlayTest/Agents/IAgent.cs

@@ -0,0 +1,11 @@
+using OECS;
+
+namespace OECS.PlayTest;
+
+/// <summary>
+/// An AI player that inspects the world and returns a decision.
+/// </summary>
+public interface IAgent<TDecision>
+{
+    TDecision Decide(World world);
+}

OECS.PlayTest/Agents/WeightedAgentPool.cs

@@ -0,0 +1,33 @@
+namespace OECS.PlayTest;
+
+/// <summary>
+/// A pool of agents selected by weight. Higher weight = more likely to be picked.
+/// </summary>
+public class WeightedAgentPool<TDecision>
+{
+    private readonly List<(IAgent<TDecision> Agent, int Weight)> _agents = new();
+    private int _totalWeight;
+    private static readonly Random _rng = new();
+
+    public void Add(IAgent<TDecision> agent, int weight)
+    {
+        _agents.Add((agent, weight));
+        _totalWeight += weight;
+    }
+
+    public IAgent<TDecision> Pick()
+    {
+        if (_agents.Count == 0)
+            throw new InvalidOperationException("WeightedAgentPool is empty");
+
+        int roll = _rng.Next(_totalWeight);
+        int cumulative = 0;
+        foreach (var (agent, weight) in _agents)
+        {
+            cumulative += weight;
+            if (roll < cumulative)
+                return agent;
+        }
+        return _agents[^1].Agent;
+    }
+}