Threadpool for "vector in, vector out" mode

[tirimatangi]

Currently every function call is run in a separate thread. The "vector in, vector out"-mode where the same function is run over and over again with different input values might benefit from a threadpool.

Measurements show that an oldish Linux laptop running Ubuntu 20.04 can create 100000 threads per second. So, if the function call executed by the thread takes, say, 1 millisecond, the overhead from thread creation is only about 1%. Hence, the lack of threadpool is not a major issue.

[tirimatangi]

This issue was addressed in branch "thread-pool" which was merged on Nov-3-2020.