Faster version of ISA&BWT

[manpen]

Here a slightly tuned version of ISA&BWT. On my laptop, I observe a speed-up of ~320%

Questions:
In the original version, the target index of the BWT-pairs was isa-second+1. IHMO the addition is wrong, as the shift in the BWT-transform is already caused by the 1-shift between ISA and Textfile?!

Changes:

• Use STXXL-based file access for reading and writing (wont be faster, but more consistent)
• Use buffered readers and writers (probably biggest performance impact)
• Overlap writing of ISA-file and filling of BWT-sorter
• Moved from ksort to sorter which reduces the I/Os required, but does not perform an integer sort. On my laptop this causes a slow-down of ~5%. But experiments were conducted with a RAM-Disk. Hence it should be faster on secondary memory. Also, as discussed, we want to have an integer-based sorter "soon"