Fix multithreading issue in incremental cycle populator

[kpieprzak]

It is proposition of fix for issue #365

[PaulSandoz]

Thank you for doing this. I question the approach to parallelism within HollowIncrementalCyclePopulator but we can revise that later on (perhaps using parallel streams that will likely have better characteristics with regards to load distribution, and/or thinking about the axis of parallelism, such as the type state thereby not requiring a thread safe bit set).

[PaulSandoz]

IIUC running on a machine with say 32 hardware threads will potentially create a lot of threads for not much benefit as only 32 threads will be running.

Can you please reduce to say 4 (that is the degree of over-provisioning fork join does to compensate for lumpy loads) but run each test within a loop for a number of iterations, thereby increasing the likely hood that a race condition will be encountered.

[kpieprzak]

Done.

[dkoszewnik]

Thank you very much @kpieprzak