Extend EIGEN_QT_SUPPORT to Qt6

Included note on inner stride for compile-time vectors. See https://gitlab.com/libeigen/eigen/-/issues/2355#note_711078126

Fix -Wbitwise-instead-of-logical clang warning

Add nan-propagation options to matrix and array plugins.

Revert bit_cast to use memcpy for CUDA.

Use reinterpret_cast on GPU for bit_cast.

Disable Tree reduction for GPU.

Improve accuracy of full tensor reduction for half and bfloat16 by reducing leaf size in tree reduction.

Fix MSVC+NVCC EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR compilation.

Fix tuple compilation for VS2017.

Move CUDA/Complex.h to GPU/Complex.h, remove TensorReductionCuda.h

Vectorize Visitor.h.

Fix Windows CMake compiler/OS detection.

Speed up tensor reduction

Reduce tensor_contract_gpu test.