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-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(())
     }