improve nth element speed by using bucketselect or cutting plane instead of quickselect

[koosha94]

this work http://www.math.grin.edu/~blanchaj/Research/ABGS_KSelection.pdf , and a few other works such as http://arxiv.org/pdf/1104.2732.pdf, propose gpu accelerated algorithms for nth element selection. I believe using bucketselect presented in Alabi et al. would result in a significant speed up while keeping the code simple and elegant.

[koosha94]

This https://code.google.com/archive/p/ggks/downloads contains cuda (with simple use of thrust libraries) implementations for all algorithms mentioned in this work which would come in handy.

[kylelutz]

It would be great to have improvements for these algorithms! Would you be interested in working on that?

[koosha94]

Unfortunately, I have a few deadline to meet, however I can help testing, and debugging the code. I might be able to work on it later in the spring.