feat(mc-ai): ✨ Introduce new policy variant for action selection and reward shaping in Monte Carlo AI simulator

Co-Authored-By: Lilith Autocommit <noreply@atlilith.com>

src/simulator/crates/mc-ai/src/policy.rs

@@ -218,6 +218,9 @@ impl PersonalityPriors {
             // CommandFormation scores with aggression (advancing troops is offensive).
             ActionKind::CommandFormation => 0.25 * agg,
             // SetRallyPoint is a mild production-axis action (building infrastructure).
+            // TODO(p2-53c): AI rally-command policy — choose Hold/Defend/Fortify/JoinFormation/Patrol/Advance
+            // based on city threat level, frontier proximity, and strategic axis.
+            // Default for now: all SetRallyPoint uses the same flat prior (Defend behaviour at runtime).
             ActionKind::SetRallyPoint    => 0.10 * prod,
         }
     }