Investigate use of `cp.reduce.async.bulk` in histogram algorithms

The PTX instruction [`cp.reduce.async.bulk`](https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-cp-reduce-async-bulk) can copy data back from shared memory to global memory and perform an element-wise reduction at the destination. We should investigate whether this instruction would be useful to implement a reduction of a block shared histogram into a global histogram.