Implement CUDA IProductWRTBase sum-factorization kernels

CUDA kernels for IProductWRTBase operation have been implemented for all element types (seg, quad, tri, hex, tet, prism, and pyr). All implementations are based on the SIMD-based matrix-free version.

To avoid multiple copy from CPU to GPU, geometric and basis data are copied to the GPU using the constructor.

Operators/IProductWRTBase/IProductWRTBaseCUDA.cu

+#include "IProductWRTBaseCUDA.hpp"
+
+namespace Nektar::Operators::detail
+{
+template <>
+std::string OperatorIProductWRTBaseImpl<double, ImplCUDA>::className =
+    GetOperatorFactory<double>().RegisterCreatorFunction(
+        "IProductWRTBaseCUDA",
+        OperatorIProductWRTBaseImpl<double, ImplCUDA>::instantiate, "...");
+}