perf(rl): eliminate 2 of 3 per-step lobsim syncs — 21ms → 10ms/step

Remove unnecessary stream.synchronize() from:
1. step_fill_from_market_targets (lobsim): stream ordering handles
   the dependency — step_pnl_track launches on the same stream
2. step_pnl_track (lobsim): downstream kernels read pos_d via
   stream ordering, no host read needed per step

Defer pos_fraction readback by one step (same pattern as loss
readback) — eliminates the third 7ms sync. pos_fraction is a
dataset-level statistic that barely changes step to step.

nsys: 3.0 → 1.0 slow syncs/step. Sync time: 21.8 → 9.65 ms/step.
The remaining sync is the training graph event wait (actual compute).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

crates/ml-alpha/src/trainer/integrated.rs

@@ -6864,12 +6864,12 @@ impl IntegratedTrainer {
                 )
                 .context("step_with_lobsim_gpu: gather_pos_fraction")?;
 
-            // Sync to make the pos_fraction host-readable (small cost;
-            // only needed for the clamped pw computation).
-            unsafe {
-                raw_stream_sync(self.raw_stream)
-                    .map_err(|e| anyhow::anyhow!("pos_fraction sync: {:?}", e))?;
-            }
+            // One-step deferred read: the gather kernel wrote pos_fraction
+            // on the stream THIS step, but we read PREVIOUS step's data
+            // (already host-visible from prior stream work). One-step lag
+            // is acceptable — pos_fraction is a dataset-level class balance
+            // statistic that barely changes step to step. Eliminates a
+            // 7ms host-blocking stream sync.
             let pf_slice = pf_staging.read_all();
 
             // Run supervised training step.

crates/ml-backtesting/src/sim/mod.rs

@@ -1300,7 +1300,8 @@ impl LobSimCuda {
                 .arg(&n)
                 .launch(cfg)?;
         }
-        self.stream.synchronize()?;
+        // No sync needed — step_pnl_track launches on the SAME stream,
+        // so CUDA stream ordering guarantees submit_market completes first.
 
         self.step_pnl_track(current_ts_ns)?;
         Ok(())
@@ -1473,7 +1474,9 @@ impl LobSimCuda {
                 .arg(&mut self.diag_outcome_n_wins_d)
                 .launch(cfg)?;
         }
-        self.stream.synchronize()?;
+        // No sync — downstream kernels on the same stream (fused_reward_pipeline,
+        // rl_per_push) read pos_d via stream ordering. Host-side trade log reads
+        // (read_trade_records) must sync explicitly when called.
         Ok(())
     }