t-hishinuma

``` docker run -it --gpus all -v $PWD:/work nvidia/cuda:11.7.0-devel-ubuntu22.04 == apt update -y apt install -y git intel-mkl cmake ninja-build ccache clang clang-tools libomp-14-dev gcc gfortran git config --global --add...

we only use clang13 for CI document step

See https://github.com/ricosjp/monolish/issues/107

Implement transpose() on GPU

2

ref: https://docs.nvidia.com/cuda/cublas/index.html#cublas-lt-t-gt-geam

When creating a CRS class from a COO array, it is necessary to create a COO class.

but, sample code of cusparse is not updated https://docs.nvidia.com/cuda/cusparse/index.html#csric02 I dont like trial and error, so wait for the sample code to be updated.

Single precision SpMM of CUDA 11.4 did not work.

1

`test/matrix/matrix_blas.cpp` is failed

clang12.0.1 said `Unknown CUDA version. cuda.h: CUDA_VERSION=11040. Assuming the latest supported version 10.1 [-Wunknown-cuda-version]`