`@reduce` decorator

[PokhodenkoSA]

We currently do not have anything similar to @cuda.reduce and the output of this step should be a design to support a similar @reduce decorator for numba-dppy.

Features to implement:

• [ ] @dppy.reduce
• [ ] Support lambda
• [ ] Works with host and device arrays
• [ ] Optional array size parameter
• [ ] Return value or output parameter
• [ ] Optional initial value
• [ ] Optional pin to queue

Example:

@cuda.reduce
def sum_reduce(a, b):
    return a + b
res = sum_reduce(arr)

sum_reduce = cuda.reduce(lambda a, b: a + b)

[PokhodenkoSA]

Related issues:

• https://github.com/IntelPython/numba-dppy/issues/71
• https://github.com/IntelPython/numba-dppy/issues/78
• https://github.com/IntelPython/numba-dppy/issues/96
• https://github.com/IntelPython/numba-dppy/issues/93