Speedup encoding by memoizing sRGBtoLinear

[michielbuddingh]

This trades encoder performance for memory, so it
might not be desirable in all cases

[codecov-io]

Codecov Report

Merging #2 into master will increase coverage by 0.35%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master       #2      +/-   ##
==========================================
+ Coverage   92.89%   93.25%   +0.35%     
==========================================
  Files           4        4              
  Lines         169      178       +9     
==========================================
+ Hits          157      166       +9     
  Misses          7        7              
  Partials        5        5

Impacted Files	Coverage Δ
encode.go	85.54% <100%> (+1.75%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 69c0b64...67eee1b. Read the comment docs.

[michielbuddingh]

Looking at comparisons between running the old and new algorithm on the same images, there are some differences the unit tests didn't catch. Don't merge this yet, I'll get back to you on this.

[michielbuddingh]

This time it's been tested on a much larger set of images; output is identical compared to the original.

[buckket]

Having 256 additional float64s and more encoding speed seems like a good compromise. If memory constraints are an huge issue one wouldn’t use Go anyways. Thanks for your contribution!