PERF-2 P1.1: the multigrid-PCG's prepared operator (hierarchy, f64 fine level, active cells, components) cached on the embedded solver and reused while the operator is bit-identical (an exact key over the coefficient bit patterns, the mask and the hierarchy parameters); solve_multigrid_pcg_cached gives the uncached solve's answer bit for bit (pin poisson_cache_exact.rs: five right-hand sides on one operator, a one-coefficient miss, both precisions); the CG driver split into Prepared::build + run_pcg

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

crates/specialized/rtx-cfd/src/solvers/incompressible/embedded/mod.rs

@@ -171,6 +171,9 @@ pub struct EmbeddedPisoSolver {
     /// PERF-2 P0: Poisson `(setup ns, iterate ns, calls, CG iterations)`
     /// summed over every multigrid-PCG solve (`docs/perf2_campaign.md`).
     poisson_profile: std::cell::Cell<(u64, u64, u64, u64)>,
+    /// PERF-2 P1.1: the prepared multigrid operator, reused across rounds
+    /// and steps while the operator's coefficients and mask are unchanged.
+    pcg_cache: std::cell::RefCell<super::poisson::PcgCache>,
     moving: bool,
     /// Mask hysteresis band in multiples of the min cell size (0 = off).
     mask_hysteresis: f64,
@@ -198,6 +201,7 @@ impl EmbeddedPisoSolver {
             fringe_correction: Vec::new(),
             inner_stop_factor: 1e-2,
             poisson_profile: std::cell::Cell::new((0, 0, 0, 0)),
+            pcg_cache: std::cell::RefCell::new(super::poisson::PcgCache::default()),
             moving: false,
             mask_hysteresis: 0.0,
             time: 0.0,