demosaic: rewrite ARI to paper-exact, optimize; honest about outcome

The previous ARI implementation had quality issues (color fringing,
maze patterns on mire1) because it cut too many corners in the
adaptive selection heuristic. This rewrites ARI to match the Monno
2015 paper line-by-line (structurally equivalent to the authors'
MATLAB reference, within 0.002 dB CPSNR of a Python port).

Quality is indeed better:
- Kodak low-ISO CPSNR 39.94 dB, +0.8 dB over amaze
- Chroma zipper smallest across all tested methods and ISOs
- SIDD real raw iso_clean: +0.38 dB over amaze+dual

However, the paper-exact pipeline is slow:
- 30-80x slower than existing methods
- 53s per 15 MP patch at 16 threads (standalone)
- 113s through darktable pipeline on mire1.cr2
- Memory-bandwidth bound; OpenCL is unlikely to reach practical speeds

Optimizations applied (all preserve paper-exact output):
- Rolling sum box filter replacing double integral image
- Row-major vertical pass for cache-line reuse on wide images
- Paired GF sum sharing (~33 % box-call reduction)
- Separable Gaussian 5x5 for criterion smoothing
- Separable bicubic 7x7 for R/B residual upsample
- Removed redundant refined-estimate recomputation in RI-H
- Criterion-side adjacent-loop fusion

Menon (2007) is separately included as a reference implementation.
It is fast (~amaze speed), but benchmarking across Kodak-24 +
SIDD-medium did not show a clear advantage over amaze - chroma
zipper is similar or worse (especially cz_peak), and CPSNR is
within 0.1 dB.

Both algorithms looked strong on paper; extensive benchmarking
showed neither clearly beats the existing methods across the
tested conditions. I'll leave the merge decision to maintainers.
The benchmark numbers themselves should be useful regardless;
full writeup + reproducible scripts + Python reference ported
from the original MATLAB are available in the sibling
darktable-rawforge/demosaic_test/ directory (BENCHMARK.md).

menon.c: keep the unused FIR/k_b locals (they're referenced via
OpenMP firstprivate clauses) to fix -Werror=unused-variable with
GCC 15.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: dregsist

src/iop/demosaic.c

@@ -47,7 +47,7 @@
 #include <string.h>
 #include <time.h>
 
-DT_MODULE_INTROSPECTION(7, dt_iop_demosaic_params_t)
+DT_MODULE_INTROSPECTION(8, dt_iop_demosaic_params_t)
 
 #define DT_DEMOSAIC_XTRANS 1024 // masks for non-Bayer demosaic ops
 #define DT_DEMOSAIC_DUAL 2048   // masks for dual demosaicing methods
@@ -168,6 +168,7 @@ typedef struct dt_iop_demosaic_params_t
   float cs_center;                              // $MIN: 0.0 $MAX: 1.0 $DEFAULT: 0.0 $DESCRIPTION: "sharp center"
   gboolean cs_enabled;                          // $DEFAULT: FALSE $DESCRIPTION: "capture sharpen"
   dt_iop_demosaic_ari_quality_t ari_quality;    // $DEFAULT: DT_ARI_QUALITY_BALANCED $DESCRIPTION: "ARI quality"
+  gboolean menon_refine;                        // $DEFAULT: TRUE $DESCRIPTION: "Menon refining step"
 } dt_iop_demosaic_params_t;
 
 typedef struct dt_iop_demosaic_gui_data_t
@@ -182,6 +183,7 @@ typedef struct dt_iop_demosaic_gui_data_t
   GtkWidget *dual_thrs;
   GtkWidget *lmmse_refine;
   GtkWidget *ari_quality;
+  GtkWidget *menon_refine;
   GtkWidget *cs_thrs;
   GtkWidget *cs_radius;
   GtkWidget *cs_boost;
@@ -274,6 +276,7 @@ typedef struct dt_iop_demosaic_data_t
   float cs_center;
   gboolean cs_enabled;
   dt_iop_demosaic_ari_quality_t ari_quality;
+  gboolean menon_refine;
 } dt_iop_demosaic_data_t;
 
 static gboolean _get_thumb_quality(const int width, const int height)
@@ -481,19 +484,49 @@ int legacy_params(dt_iop_module_t *self,
    return 0;
   }
 
+  typedef struct dt_iop_demosaic_params_v7_t
+  {
+    dt_iop_demosaic_greeneq_t green_eq;
+    float median_thrs;
+    dt_iop_demosaic_smooth_t color_smoothing;
+    dt_iop_demosaic_method_t demosaicing_method;
+    dt_iop_demosaic_lmmse_t lmmse_refine;
+    float dual_thrs;
+    float cs_radius;
+    float cs_thrs;
+    float cs_boost;
+    int cs_iter;
+    float cs_center;
+    gboolean cs_enabled;
+    dt_iop_demosaic_ari_quality_t ari_quality;
+  } dt_iop_demosaic_params_v7_t;
+
   if(old_version == 6)
   {
     const dt_iop_demosaic_params_v6_t *o = (dt_iop_demosaic_params_v6_t *)old_params;
-    dt_iop_demosaic_params_t *n = malloc(sizeof(dt_iop_demosaic_params_t));
+    dt_iop_demosaic_params_v7_t *n = malloc(sizeof(dt_iop_demosaic_params_v7_t));
     memcpy(n, o, sizeof *o);
     n->ari_quality = DT_ARI_QUALITY_BALANCED;
 
     *new_params = n;
-    *new_params_size = sizeof(dt_iop_demosaic_params_t);
+    *new_params_size = sizeof(dt_iop_demosaic_params_v7_t);
     *new_version = 7;
     return 0;
   }
 
+  if(old_version == 7)
+  {
+    const dt_iop_demosaic_params_v7_t *o = (dt_iop_demosaic_params_v7_t *)old_params;
+    dt_iop_demosaic_params_t *n = malloc(sizeof(dt_iop_demosaic_params_t));
+    memcpy(n, o, sizeof *o);
+    n->menon_refine = TRUE; // default: always refine (preserves previous behavior)
+
+    *new_params = n;
+    *new_params_size = sizeof(dt_iop_demosaic_params_t);
+    *new_version = 8;
+    return 0;
+  }
+
   return 1;
 }
 
@@ -906,7 +939,7 @@ void process(dt_iop_module_t *self,
           }
         }
         else if(method == DT_IOP_DEMOSAIC_MENON)
-          menon_demosaic(t_out, t_in, width, t_rows, filters);
+          menon_demosaic(t_out, t_in, width, t_rows, filters, d->menon_refine);
         else if(method == DT_IOP_DEMOSAIC_ARI)
           ari_demosaic(t_out, t_in, width, t_rows, filters, d->ari_quality);
         else if(method == DT_IOP_DEMOSAIC_RCD)
@@ -1420,6 +1453,7 @@ void commit_params(dt_iop_module_t *self,
   d->dual_thrs = p->dual_thrs;
   d->lmmse_refine = p->lmmse_refine;
   d->ari_quality = p->ari_quality;
+  d->menon_refine = p->menon_refine;
   dt_iop_demosaic_method_t use_method = p->demosaicing_method;
   d->cs_radius = p->cs_radius;
   d->cs_thrs = p->cs_thrs;
@@ -1634,6 +1668,7 @@ void gui_changed(dt_iop_module_t *self, GtkWidget *w, void *previous)
   gtk_widget_set_visible(g->dual_thrs, isdual);
   gtk_widget_set_visible(g->lmmse_refine, islmmse);
   gtk_widget_set_visible(g->ari_quality, use_method == DT_IOP_DEMOSAIC_ARI);
+  gtk_widget_set_visible(g->menon_refine, use_method == DT_IOP_DEMOSAIC_MENON);
 
   const gboolean monomode = use_method == DT_IOP_DEMOSAIC_PASSTHROUGH_MONOCHROME
                         ||  use_method == DT_IOP_DEMOSAIC_PASSTHR_MONOX;
@@ -1830,6 +1865,11 @@ void gui_init(dt_iop_module_t *self)
       "balanced: MLRI + guided-filter RI adaptive selection\n"
       "quality: full 3-candidate adaptive selection"));
 
+  g->menon_refine = dt_bauhaus_toggle_from_params(self, "menon_refine");
+  gtk_widget_set_tooltip_text(g->menon_refine,
+    _("Menon refining step: iterative refinement of G, R, and B channels.\n"
+      "Improves quality at the cost of slightly more processing time."));
+
   g->color_smoothing = dt_bauhaus_combobox_from_params(self, "color_smoothing");
   gtk_widget_set_tooltip_text(g->color_smoothing, _("how many color smoothing median steps after demosaicing"));