Refinement: time per stage - an NVTX range per call and one TIMING line per stage with its sum and calls

RefinementStage sums the time of each stage over a refinement (Refit and its parts, the evaluations and
their X-ray, restraint, host and upload shares, the repulsion build, the riding hydrogens, the rungs) and
logs it at debug level at the end; rugnux_refine_test -v shows it. No change to any result.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015CeEGgpKdfkCTJb3YNMJcu
(cherry picked from commit 9491f54e0beccf36c65ebcec9afff90d08485f86)
(cherry picked from commit 0ce206a88df960754476ce2f75ffaed90fbf3123)

image_analysis/structure_refinement/RefinementGPU.cpp

@@ -5,6 +5,7 @@
 
 #include <algorithm>
 #include <cmath>
+#include <optional>
 
 #include "gemmi/dencalc.hpp"
 #include "gemmi/it92.hpp"
@@ -194,7 +195,11 @@ public:
     double Evaluate(const gemmi::Model &model, std::vector<std::array<double, 4>> *grad) override {
         SetAtoms(model);
         std::vector<std::array<float, 4>> g;
-        const double value = engine_->Evaluate(grad != nullptr ? &g : nullptr);
+        double value = 0;
+        {
+            const RefinementStage stage("gpu.evaluate");
+            value = engine_->Evaluate(grad != nullptr ? &g : nullptr);
+        }
         if (grad != nullptr) {
             grad->resize(g.size());
             for (size_t i = 0; i < g.size(); i++)
@@ -213,7 +218,11 @@ public:
         for (size_t i = 0; i < xyzb.size(); i++)
             x[i] = {static_cast<float>(xyzb[i][0]), static_cast<float>(xyzb[i][1]), static_cast<float>(xyzb[i][2]),
                     static_cast<float>(xyzb[i][3])};
-        const double value = engine_->EvaluateRestraints(which, x, grad != nullptr ? &g : nullptr);
+        double value = 0;
+        {
+            const RefinementStage stage("gpu.restraints");
+            value = engine_->EvaluateRestraints(which, x, grad != nullptr ? &g : nullptr);
+        }
         if (grad != nullptr) {
             grad->resize(g.size());
             for (size_t i = 0; i < g.size(); i++)
@@ -235,6 +244,7 @@ private:
 
     // The atoms' densities (made again only for the atoms whose B moved) and grid positions, to the device.
     void SetAtoms(const gemmi::Model &model) {
+        std::optional<RefinementStage> stage("gpu.atoms_host");
         const std::vector<const gemmi::Atom *> atoms = Atoms(model);
         if (last_b_.size() != atoms.size()) {
             last_b_.assign(atoms.size(), NAN);
@@ -250,6 +260,8 @@ private:
             const gemmi::Fractional f = cell_.fractionalize(atoms[i]->pos);
             pos[i] = {static_cast<float>(f.x * n3[0]), static_cast<float>(f.y * n3[1]), static_cast<float>(f.z * n3[2]), 0};
         }
+        stage.reset();
+        const RefinementStage upload("gpu.upload");
         engine_->SetAtoms(atoms_, pos);
     }