PERF-2 P3-ii: the device V-cycle as the CG's preconditioner — poisson/device.rs (one CUDA runtime per process, persistent per-operator buffers, the mg_vcycle.cu kernels at K = 1; upload r, run the V-cycle, download z; the f64 CG unchanged), MultigridParameters::device, Prepared holds the device hierarchy, the CG driver destructures the prepared operator instead of cloning it; EmbeddedPisoSolver::set_poisson_device, overset pass-through, harness knob RTX_FSI2O_MG_DEVICE=1; export_levels factored out; the quarantine's dangling cfg attribute fixed

Co-Authored-By: Claude Fable 5.1 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YJPeT6WA2e7YvAnS875AHL

crates/specialized/rtx-fsi/tests/fsi2_harness/overset.rs

@@ -326,6 +326,14 @@ impl OversetFluid {
             background.set_poisson_threads(threads);
             println!("  multigrid threads: {threads} (RTX_THREADS)");
         }
+        // PERF-2 P3-ii: `RTX_FSI2O_MG_DEVICE=1` — the V-cycle on the CUDA
+        // device (needs a `cuda`-feature build and the red-black smoother).
+        if std::env::var("RTX_FSI2O_MG_DEVICE").is_ok_and(|v| v == "1") {
+            background.set_poisson_device(true);
+            println!(
+                "  multigrid V-cycle: CUDA DEVICE, f32 red-black (PERF-2 regime, RTX_FSI2O_MG_DEVICE)"
+            );
+        }
         background.set_boundary_velocity(move |x, y, t| {
             if x <= 0.0 {
                 (inflow_for(u_mean, y, t), 0.0)