Port the SAR3D background gate, CLEAN and object boxes from Multi_depthnorm_clean

- BG_GATE_ENABLE: frame peak over its median background (sar3d_background); the
  GUI scales the frame with bg_gate_gain and its cloud threshold.
- SAR_CLEAN_ENABLE: CLEAN of a one-frame aperture on a 4x10x4 cm search grid over
  the whole scene; objects in their own scale (sar3d_objects).
- SAR_MARK_ENABLE: boxes by class around object groups (sar3d_marks), with the
  legend under the views.
- SAR_VOXEL_M as three steps X/Y/Z with the notebook defaults 0.025/0.02/0.02.
- GPU backend: image/download/load/project API, per-plane block reduction for the
  maximum, depth gate without FMA; sar3d host code built with -ffp-contract=off.
- References regenerated from notebook cells 1 and 5, with new CLEAN scenes and a
  synthetic box-rule case; docs updated.

Makefile

@@ -25,6 +25,7 @@ endif
 # Rewritten only when the CUDA setup changes, so installing (or removing) CUDA
 # rebuilds the sar3d GPU code instead of keeping a stale build.
 SAR3D_CUDA_STAMP := $(BUILD_DIR)/sar3d_cuda.flags
+SAR3D_HOST_FLAGS := -ffp-contract=off
 $(shell mkdir -p $(BUILD_DIR) && echo '$(SAR3D_CXXFLAGS) $(SAR3D_LDFLAGS)' | cmp -s - $(SAR3D_CUDA_STAMP) || echo '$(SAR3D_CXXFLAGS) $(SAR3D_LDFLAGS)' > $(SAR3D_CUDA_STAMP))
 
 INCLUDES := \
@@ -57,6 +58,7 @@ PROCESSOR_SOURCES := \
 	data_acq_and_processing/processing/processors/src/gpr_processor.cpp \
 	data_acq_and_processing/processing/processors/src/passthrough_processor.cpp \
 	data_acq_and_processing/processing/processors/src/sar3d_processor.cpp \
+	data_acq_and_processing/processing/processors/src/sar3d_clean.cpp \
 	data_acq_and_processing/processing/processors/src/sar3d_gpu.cpp \
 	data_acq_and_processing/processing/data_processor/src/processing_live_config.cpp \
 	data_acq_and_processing/processing/data_processor/src/data_processor.cpp \
@@ -122,13 +124,14 @@ check-native: $(NATIVE_BUILD_DIR)/native_regressions $(BIN_DIR)/collector_regres
 BACKPROJECTION_TEST_SOURCES := data_acq_and_processing/tests/backprojection_fixture.cpp \
 	data_acq_and_processing/common_cpp/ipc/src/shared_types.cpp \
 	data_acq_and_processing/processing/processors/src/sar3d_processor.cpp \
+	data_acq_and_processing/processing/processors/src/sar3d_clean.cpp \
 	data_acq_and_processing/processing/processors/src/sar3d_gpu.cpp \
 	data_acq_and_processing/processing/processors/src/gpr_processor.cpp
 BACKPROJECTION_TEST_OBJS := $(addprefix $(NATIVE_BUILD_DIR)/,$(BACKPROJECTION_TEST_SOURCES:.cpp=.o))
 DEPFILES += $(BACKPROJECTION_TEST_OBJS:.o=.d)
 
-# The sar3d processor on the real CUDA backend against the notebook reference;
-# without a usable GPU the test checks the warning and skips the rest.
+# The sar3d processor on the real CUDA backend against the notebook reference; without a
+# usable GPU the test checks the warning and the box rules and skips the rest.
 check-sar3d: $(NATIVE_BUILD_DIR)/backprojection_fixture
 	BACKPROJECTION_FIXTURE="$(abspath $<)" $(PYTHON) data_acq_and_processing/tests/test_backprojection.py Sar3dTest
 
@@ -139,6 +142,7 @@ $(NATIVE_BUILD_DIR)/backprojection_fixture: $(BACKPROJECTION_TEST_OBJS)
 	$(CXX) $(NATIVE_CXXFLAGS) $^ -o $@ $(LDFLAGS) $(SAR3D_LDFLAGS)
 
 $(NATIVE_BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_gpu.o: NATIVE_CXXFLAGS += $(SAR3D_CXXFLAGS)
+$(NATIVE_BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_%.o: NATIVE_CXXFLAGS += $(SAR3D_HOST_FLAGS)
 $(NATIVE_BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_gpu.o: $(SAR3D_CUDA_STAMP)
 
 $(NATIVE_BUILD_DIR)/native_regressions: $(NATIVE_TEST_OBJS)
@@ -159,6 +163,9 @@ $(BIN_DIR)/data_processor: $(DATA_PROCESSOR_OBJS) $(SAR3D_CUDA_STAMP)
 
 $(BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_gpu.o: CXXFLAGS += $(SAR3D_CXXFLAGS)
 $(BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_gpu.o: $(SAR3D_CUDA_STAMP)
+# The sar3d host code breaks the same ties as NumPy (grid corners on the depth gate, range
+# rings), so it is not contracted into fused multiply-adds where the target has them.
+$(BUILD_DIR)/data_acq_and_processing/processing/processors/src/sar3d_%.o: CXXFLAGS += $(SAR3D_HOST_FLAGS)
 
 # Keep source-relative paths so moving a source also replaces its dependencies.
 $(RPH_BUILD_DIR)/%.o: %.cpp