CUDA device synchronization is performed only for CUDA benchmarks.

src/Benchmarks/BLAS/array-operations.h

src/Benchmarks/BLAS/spmv.h

src/Benchmarks/BLAS/vector-operations.h

src/Benchmarks/Benchmarks.h

src/Benchmarks/DistSpMV/tnl-benchmark-distributed-spmv.h

src/Benchmarks/LinearSolvers/benchmarks.h

src/Benchmarks/Traversers/tnl-benchmark-traversers.h