Check for CUDA errors after running benchmark

[alliepiper]

We recently had an issue where a benchmark kernel caused an illegal memory access, and the error was asynchronously reported in an unrelated NVBench CUDA API call. Any errors emitted during the execution of a benchmark should be correctly reported as originating from the benchmark execution.

We should add NVBENCH_CUDA_CALL(cudaGetLastError()); or similar in the detail/measure_* runners after each kernel execution.