gpairs benchmark: dpnp on CPU is 2.5 times slower than NumPy

[antonwolfy]

Results from dpbench of gpairs workload (running on a laptop with 11th Gen processor and Iris Xe graphics):

workload	numpy	dpnp CPU	dpnp GPU	size
gpairs	6.6 s	15.8 s	8.6 s	8192

demonstrates dpnp execution time on CPU which is 2.5 times greater than one of NumPy.

[antonwolfy]

The numbers with dpnp=0.12.0:

workload	numpy	dpnp CPU	dpnp GPU	size
gpairs	5.8 s	12.3 s	8.3 s	8192

The result is faster, but still not in the target.