Use PROOF to speed up cutapplier

[admorris]

I noticed this mentioned on the Mattermost (Example). Doesn't seem too complicated to implement but I don't have time immediately, so I'm making this issue to remind myself.

[admorris]

It's more complicated than I thought. Apparently you have to write a whole TSelector class just to apply a cut in parallel?

[cofitzpa]

I always assumed the bottleneck for cutapplier was the I/O rather than the CPU, so never investigated using PROOF for it. The place where parallelisation would be handy is in crop and bwdiv imo. I've already parallelised bwdiv in a private version using TProcessExecutor and that works very nicely.