feat: add counterfactual regret minimization to agents

Summary:
- Add CFR game state tree that uses a vec as an allocation arena.
- Add a historian that watches actions and moves the cfr tree along
- Add an agent that uses the tree to create new actions.

Test Plan:
- Added a test.

Cargo.toml

@@ -19,6 +19,8 @@ arbitrary = { version = "1.3.2", optional = true, features = ["derive"] }
 tracing = { version = "0.1.40", optional = true}
 approx = { version = "0.5.1", optional = true}
 uuid  = {version = "1.8.0", optional = true, features = ["v7"]}
+little-sorry = { version = "0.4.0", optional = true}
+ndarray = { version = "0.15.6", optional = true}
 
 [dev-dependencies]
 criterion = "0.5.1"
@@ -32,7 +34,7 @@ approx = { version = "0.5.1"}
 default = ["arena", "serde"]
 uuid = ["dep:uuid"]
 serde = ["dep:serde", "uuid?/serde"]
-arena = ["dep:tracing", "uuid"]
+arena = ["dep:tracing", "dep:little-sorry", "dep:ndarray", "uuid"]
 arena-test-util = ["arena", "dep:approx"]
 
 [[bench]]