use an rgba Pix array instead of converting each x,y coord to a Color

[hunterloftis]

An illustration of how you might make a more apples-to-apples comparison, with quite different performance characteristics.

~/code/src/github.com/nicolashahn/diffimg-go (master) 🐵  go build ./main.go && time ./main ~/Desktop/test-image-1.png ~/Desktop/test-image-2.png
0.12764707841506104

real	0m3.291s
user	0m3.100s
sys	0m0.197s
~/code/src/github.com/nicolashahn/diffimg-go (master) 🐬  git checkout faster-array
Switched to branch 'faster-array'
Your branch is up to date with 'origin/faster-array'.
~/code/src/github.com/nicolashahn/diffimg-go (faster-array) 🐰  go build ./main.go && time ./main ~/Desktop/test-image-1.png ~/Desktop/test-image-2.png
0.12764707841506104

real	0m1.499s
user	0m1.386s
sys	0m0.131s

[hunterloftis]

I noticed in the writeup that the Rust version was just iterating through a pixel array while the Go version was iterating through nested X, Y, color-channel values, converting each pixel to a Color, then back to an array of uints. This shows how you can just iterate through the pixels without all that extra work for a significant speedup.