Scale: measure time and memory on projects of 30k to 100k lines #69

Measure and, where needed, fix how the analyzer scales; the largest repository of the regression corpus has 213 functions.

• Analyse two or three open-source projects of 30 000 to 100 000 lines and record time, peak memory, coverage and finding counts.
• Measure the effect of --cache on a second run and of --jobs on a cold run at that size.
• Profile and fix the hot spots or the memory retention found, if any.
• Record the results in the documentation and add the largest project that stays fast enough to the regression corpus.