feat: log full validation backtest metrics — Sharpe + Sortino + WinRate + MaxDD + Trades + Calmar

compute_validation_loss only returned Sharpe, discarding all other
backtest metrics. Now logs the complete picture on every epoch so we
can verify if the learning system actually works on out-of-sample data.

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

crates/ml/src/trainers/dqn/trainer/metrics.rs

@@ -554,10 +554,16 @@ impl DQNTrainer {
             )
         }.map_err(|e| anyhow::anyhow!("GPU backtest evaluate_dqn_graphed: {e}"))?;
 
-        // Single window — take its Sharpe directly
-        let val_sharpe = metrics.first()
-            .map(|m| m.sharpe as f64)
-            .unwrap_or(0.0);
+        // Single window — log full validation metrics
+        let val_sharpe = if let Some(m) = metrics.first() {
+            tracing::info!(
+                "Validation backtest: Sharpe={:.2} Sortino={:.2} WinRate={:.1}% MaxDD={:.3}% Trades={:.0} Calmar={:.2}",
+                m.sharpe, m.sortino, m.win_rate * 100.0, m.max_drawdown * 100.0, m.total_trades, m.calmar,
+            );
+            m.sharpe as f64
+        } else {
+            0.0
+        };
 
         Ok(-val_sharpe)
     }