Parallel implementation of map

[VlasovRoman via Digitalmars-d-learn]

Hey, guys!
After many attempts pmap function was realized by me. This is 
analogous to taskPool.amap function, but I do not know how much 
he is good. In simple tests, it showed a corresponding 
acceleration. I would like to ask you to advise me options to 
improve this function. If all this makes sense, I can make a DUB 
package.

https://gist.github.com/VlasovRoman/4c78a123acd44b3da8b96b43637df080