CLAUDE: Step 3 — backend selector (GpuQuad.create factory + -Dtp.backend=jna), 32 sites routed

GpuQuad.create(gpuTileProcessor, quadCLT, debug) returns the JCuda GpuQuad by default, or the native
GpuQuadJna when -Dtp.backend=jna (srcdir/devrt overridable via -Dtp.jna.srcdir / -Dtp.jna.devrt).
Routed all 32 main+aux `new GpuQuad(...)` 3-arg sites in Eyesis_Correction.java through the factory.
JCUDA remains the default (behavior identical when the property is unset). mvn -DskipTests compile clean.

Migration now fully implemented + compiling end-to-end (Step 1 native TpProc API, Step 2 GpuQuadJna
full CUAS surface, Step 3 selector). Ready for the JCUDA-vs-JNA comparison + incremental troubleshooting.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

src/main/java/com/elphel/imagej/gpu/GpuQuad.java

@@ -490,6 +490,25 @@ public class GpuQuad{ // quad camera description
 		texture_stride_rgba = (int)(device_stride[0] / Sizeof.FLOAT);
 	}
 	
+	// Backend selector (architecture B). Default JCUDA; set -Dtp.backend=jna to use the native
+	// (libtileproc.so via JNA) backend GpuQuadJna instead. JNA mode never initializes JCuda.
+	// Validate by running the same workflow both ways and diffing saved outputs.
+	public static boolean useJnaBackend() {
+		return "jna".equalsIgnoreCase(System.getProperty("tp.backend", "jcuda"));
+	}
+	public static GpuQuad create(
+			GPUTileProcessor gpuTileProcessor,
+			final QuadCLT    quadCLT,
+			int              debug_level) {
+		if (useJnaBackend()) {
+			String src   = System.getProperty("tp.jna.srcdir", "/home/elphel/git/tile_processor_gpu/src");
+			String devrt = System.getProperty("tp.jna.devrt",  "/usr/local/cuda/targets/x86_64-linux/lib/libcudadevrt.a");
+			System.out.println("GpuQuad.create(): NATIVE (JNA) backend for "+quadCLT+" (src="+src+")");
+			return new com.elphel.imagej.gpu.jna.GpuQuadJna(quadCLT, src, devrt, debug_level);
+		}
+		return new GpuQuad(gpuTileProcessor, quadCLT, debug_level);
+	}
+
 	// No-allocation constructor for the native (JNA) backend subclass GpuQuadJna.
 	// Sets the final config fields from quadCLT but allocates NO JCuda GPU memory and creates NO
 	// JCuda context (gpuTileProcessor = null). The subclass owns its native (TpProc) GPU memory and