Support calling algorithms for partitions

[kordejong]

Slope is implemented in terms of other algorithms (convolve, divide, pow, sqrt). When slope is called for parallel execution, each of these algorithms represents a synchronisation point. It would be better to have slope call itself for each partition that is calculated in parallel, and have each sub-slope call execute until the end. Then the end of the slope call itself will be a single synchronisation point, which is an improvement.

This work is necessary to get rid of synchronization points in general, when combining algorithms.