fix: diversity penalty max_entropy — log2(3) for 3-bucket input, not log2(9)

The stale comments agent incorrectly changed calculate_diversity_penalty's
max_entropy from log2(3) to log2(9). But this function takes 3-bucket
BUY/SELL/HOLD ratios — max entropy of 3 buckets IS log2(3).

The log2(9) change was correct for the LOGGING in extract_objective
(which displays max_entropy for 9 exposure actions), but wrong for the
penalty function that operates on 3 action categories.

Reverted to log2(3) and restored test assertions.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

crates/ml/src/hyperopt/adapters/dqn.rs

@@ -2084,7 +2084,7 @@ fn calculate_diversity_penalty(action_distribution: &[f64; 3]) -> f64 {
         .collect();
 
     let entropy = calculate_action_entropy(&*action_counts);
-    let max_entropy = (9.0_f64).log2(); // ~3.170
+    let max_entropy = (3.0_f64).log2(); // ~1.585 (3 buckets: BUY/SELL/HOLD)
 
     // Smooth quadratic penalty: -5.0 * (1 - entropy/max_entropy)²
     // At entropy=0: -5.0, at entropy=max: 0.0
@@ -4237,12 +4237,12 @@ mod tests {
         assert!(penalty_mid > -1.0, "Medium entropy mild: {}", penalty_mid);
 
         let penalty_uniform = calculate_diversity_penalty(&[0.33, 0.33, 0.34]);
-        assert!(penalty_uniform.abs() < 0.1, "Uniform ~0: {}", penalty_uniform);
+        assert!(penalty_uniform.abs() < 0.1, "Uniform 3-bucket ~0: {}", penalty_uniform);
 
-        // Monotonic
-        assert!(penalty_uniform > penalty_mid);
-        assert!(penalty_mid > penalty_low);
-        assert!(penalty_low > penalty_zero);
+        // Monotonic: uniform > mid > low > zero (less negative = better)
+        assert!(penalty_uniform > penalty_mid, "uniform > mid: {} > {}", penalty_uniform, penalty_mid);
+        assert!(penalty_mid > penalty_low, "mid > low: {} > {}", penalty_mid, penalty_low);
+        assert!(penalty_low > penalty_zero, "low > zero: {} > {}", penalty_low, penalty_zero);
     }
 
     #[test]