Stream in the context-manager form is not used in `ElementwiseKernel` or `ReductionKernel`

This is actually a bug reported back in #1695 that unfortunately went unnoticed.

In examples/stream/map_reduce.py, a list of streams was created for executing cupy.matmul() in parallel, which is backed by a ReductionKernel in this case: https://github.com/cupy/cupy/blob/1af22f57fda92ae35bde806d0c4d110faf4fed52/cupy/core/core.pyx#L2513-L2516 However, inspecting the implementation I found that ReductionKernel only accepts an explicit stream argument; it does not pick up any current stream: https://github.com/cupy/cupy/blob/32718607a7808ec6bc3a24cf9231a9351f8fc95e/cupy/core/reduction.pxi#L396 In other words, that example was misleading because those streams were not used at all and so all executions were serialized, as can be checked from nvprof + nvvp (see the circle in red):

The same bug also appears in ElementwiseKernel: https://github.com/cupy/cupy/blob/1af22f57fda92ae35bde806d0c4d110faf4fed52/cupy/core/_kernel.pyx#L537

In my opinion, unlike RawKernel which is not used by any CuPy core functionalities, ElementwiseKernel and ReductionKernel should honor the current stream by checking the current stream pointer if no stream argument is explicitly given, since many CuPy functions like cupy.matmul() do not support passing in a stream. A similar approach is already adopted in the FFT module, see #2362.

[emcastillo]

Actually, the previous code was fine, just increasing the matrix sizes shows real overlap

[emcastillo]

Apparently, if the stream is set to None in those two functions, when the kernel is launched the current stream is retrieved: https://github.com/cupy/cupy/blob/32718607a7808ec6bc3a24cf9231a9351f8fc95e/cupy/cuda/function.pyx#L126

This is done in the linear_launch function https://github.com/cupy/cupy/blob/32718607a7808ec6bc3a24cf9231a9351f8fc95e/cupy/cuda/function.pyx#L174