parallelize block splitting

[mmolari]

parallelize block splitting during reweave

[ivan-aksamentov]

@mmolari This looks alright technically, but in my measurements has no effect at all on running time of the program, just like the changes in #108.

Unless you insist on this, I propose we convert it to draft PR and wait for better days when we figure out why this happens.

It might be that this place is simply not a bottleneck, and no matter how you speedup a 0.1% portion of total run time, even if we make this code complete in 0 ms, the program still runs in the same time overall. Or it could be that any speedup is erased by the threading overhead (e.g. data sharing between threads). We need some more research and probably a parallel traversal.

I leave the final decision to you.

[mmolari]

I agree, you're right. Let's leave this open for now and think about this later, it's just a marginal improvement...