Refactor NSS and DNSS implementations for improved readability and modularity

- Rewrote legacy code into clean C++17 style.
- Added missing method implementations and input validation.
- Introduced options struct for configurable parameters in NSS and DNSS.
- Implemented CUDA support for DNSS rotational feature computation.
- Added new CMake configuration for building with or without CUDA.
- Updated README to reflect changes and provide build instructions.
- Added .gitignore to exclude build directories.

Author: eclipse0922

.gitignore

@@ -0,0 +1,2 @@
+build/
+build-cuda/

CMakeLists.txt

@@ -0,0 +1,21 @@
+cmake_minimum_required(VERSION 3.20)
+
+project(DNSS LANGUAGES CXX)
+
+option(DNSS_ENABLE_CUDA "Enable CUDA acceleration for DNSS rotational feature computation" OFF)
+
+add_library(dnss NSS.cpp NSS.h)
+target_include_directories(dnss PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_compile_features(dnss PUBLIC cxx_std_17)
+
+if(DNSS_ENABLE_CUDA)
+	enable_language(CUDA)
+	find_package(CUDAToolkit REQUIRED)
+	if(NOT CMAKE_CUDA_ARCHITECTURES)
+		set(CMAKE_CUDA_ARCHITECTURES native)
+	endif()
+	target_sources(dnss PRIVATE dnss_cuda.cu)
+	target_compile_definitions(dnss PRIVATE DNSS_HAS_CUDA=1)
+	target_link_libraries(dnss PUBLIC CUDA::cudart)
+	set_target_properties(dnss PROPERTIES CUDA_SEPARABLE_COMPILATION ON CUDA_RESOLVE_DEVICE_SYMBOLS ON)
+endif()