Investigate pSACAK for parallel suffix sorting

[lberrymage]

Suffix sorting takes up a majority significant portion of diffing time (the only more significant being compression). We currently use the SACA-K algorithm to perform this step, which works fine. However, it is strictly single-threaded, severely limiting the speed of diffing.

Another suffix sorting algorithm partially by the same author exists called pSACAK, which is essentially a parallelizable version of SACA-K. We should investigate the feasibility of implementing pSACAK and using it in our suffix sorting library instead of SACA-K for the significant performance gains it theoretically provides.

[lberrymage]

Further improvements have been made to pSACAK since its original publishing in the following papers:

• In-Place Suffix Sorting on a Multicore Computer with Better Design, which uses a new naming procedure to increase performance over pSACAK by approximately 20% (creating pSACAK+).
• Scalable Suffix Sorting on a Multicore Machine, which describes a new algorithm, uSAIS, which has has a speedup over pSACAK of ~1.19 and can be adapted to make use of external memory when there isn't enough internal memory available.

It may be possible for the pSACAK+ naming improvements to be incorporated into uSAIS, although I haven't done thorough enough research into either to see whether this is possible with the more extensive pipelining present in uSAIS.