Add NvJpegPool (4 encoders) and JPEG passthrough in BmpToJpeg

- NvJpegPool: singleton pool of 4 NvJpegCompressor instances with
  lock-free slot acquisition (~160MB VRAM). Threads that can't grab
  a slot fall back to TurboJPEG with zero wait.
- JPEG passthrough: BmpToJpeg now checks if input is already JPEG
  (FF D8 FF magic) and copies directly without re-encoding.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

modules/ANSCV/ANSOpenCV.h

@@ -5,6 +5,9 @@
 #include "ANSLicense.h"
 #include "LabVIEWHeader/extcode.h"
 #include <vector>
+#include <array>
+#include <atomic>
+#include <memory>
 #include <opencv2/opencv.hpp>
 
 // Forward declaration for NI Vision IMAQ Image (avoids nivision.h dependency for consumers)
@@ -63,6 +66,26 @@ namespace ANSCENTER
 		unsigned char*  _gpuBuffer = nullptr;     // reusable device memory
 		size_t          _gpuBufferSize = 0;
 	};
+
+	// Fixed-size pool of NvJpegCompressors (~40MB VRAM each).
+	// Threads that can't acquire an encoder fall back to TurboJPEG.
+	class NvJpegPool {
+	public:
+		static constexpr int kPoolSize = 4;
+		static NvJpegPool& Instance();
+		// Try to compress with nvJPEG. Returns empty string if no encoder
+		// available or on non-NVIDIA hardware — caller should fall back.
+		[[nodiscard]] std::string tryCompress(const cv::Mat& image, int quality);
+		[[nodiscard]] bool isAvailable() const noexcept { return _available; }
+	private:
+		NvJpegPool();
+		~NvJpegPool() = default;
+		NvJpegPool(const NvJpegPool&) = delete;
+		NvJpegPool& operator=(const NvJpegPool&) = delete;
+		bool                                        _available = false;
+		std::array<std::unique_ptr<NvJpegCompressor>, kPoolSize> _encoders;
+		std::array<std::atomic<bool>, kPoolSize>    _inUse;     // lock-free slot flags
+	};
     /// <summary>
 	/// // ANSOPENCV class provides various image processing functionalities using OpenCV and ANS Center SDK.
     /// </summary>