Some minor optimizations for speeding up nearest neighbor resize. (#8444)

* Some minor optimizations for speeding up nearest neighbor resize.

By removing loop over channels. Impact mainly visible for more than 3
channels. Numbers for benchmark with 6 channels:

Before:
Benchmark                                        Time(ns) Iterations
--------------------------------------------------------------------
BM_Resize_ResizeNearestNeighbor_cpu_10_499_499   53222200        100
     280.7M items/s

After:
Benchmark                                        Time(ns) Iterations
--------------------------------------------------------------------
BM_Resize_ResizeNearestNeighbor_cpu_10_499_499   44504450        100
     335.7M items/s

* Simplified implementation in accordance with jhseu@ comments.