Add stroke composite feedback planner

docs/modernization/roadmap.md

@@ -868,6 +868,13 @@ read destination color directly, while other backends must use ping-pong render
 targets backed by texture copy or render-target blit support. This is exposed
 through `pano_cli plan-paint-feedback` and tracked by DEBT-0036 until the live
 paint renderer consumes the plan.
+`pp_paint_renderer` now consumes that lower-level feedback planner through a
+stroke composite plan that decides whether a stroke/layer blend can use
+fixed-function blending or needs framebuffer-fetch/ping-pong destination
+feedback. `pano_cli plan-stroke-composite` exposes the same decision for
+automation, including layer blend, stroke blend, dual-brush, and pattern-blend
+inputs. Live canvas stroke execution is still legacy OpenGL and remains tracked
+by DEBT-0036 until the app calls through this paint-renderer boundary.
 The existing renderer classes are not yet fully
 behind the renderer interfaces.
 
@@ -1118,6 +1125,10 @@ Results:
   no-feedback blends, invalid render-target usage, unsupported backends, and
   depth-target rejection.
 - `pp_paint_renderer_compositor_tests` passed.
+  The suite now covers fixed-function stroke composite planning,
+  framebuffer-fetch planning, ping-pong texture-copy/blit fallback planning,
+  dual/pattern blend feedback detection, invalid blend mode rejection,
+  unsupported backend rejection, and invalid render-target rejection.
 - `pp_ui_core_color_tests` passed.
 - `pp_ui_core_layout_value_tests` passed.
 - `pp_ui_core_layout_xml_tests` passed.