cortex-strings benchmarks

Hi Dave. I had a little play with cortex-strings and did some benchmarks on my Tegra 2. Images are attached.

I've added two scripts to cortex-strings: scripts/bench-all.sh runs all the routines on all variants and records them scripts/plot.py plots the results from above

ploy.py corrects for the benchmark overhead by doing a linear fit to the null 'bounce' results and subtracting this fit.

You should be able to a autogen; configure; make; bash scripts/bench-all.sh | tee log.txt; python scripts/plot.py log.txt. I'm sure you have your own favourite tools though.

The string routines look good. Lumpy in funny ways though...

-- Michael