[abc] cflow with new abc

Hi all...

I've rerun the cflow benchmarks on tofu with the new abc.

With no optimizations (not even intraproc ones), abc produces bad
bytecode for lod-sim and weka. I suspect it's a method becoming too
long; the exception is:
Exception in thread "main" java.lang.ClassFormatError: weka/core/SparseInstance (Truncated class file)

On ConaSim, the interproc abc is now faster than any of the runs without
interproc, but it is still not as fast as ajc121. Here are all the times
for ComaSim:

interproc 71.702
ajc12 76.512
ajc121 69.998
all intraproc 73.834
sharing+counters 73.156
sharing 75.352
no intraproc opts 75.416

I guess I'll run things on string now. It seemed to be more favourable.

I'm attaching the full timings from tofu.

Ondrej