Part of issue509.

Once this is done, we could do issue700.

Great! I merged and will get started on issue700 now.

Experimental data looks OK to merge.

The results look OK. There again were some errors on the grid: one task never
started; there were again 10 PSR tasks where VAL ran crashed; and I got a couple
of error messages complaining about a "Read-only file system". The last one
looks like it happens when slurm is cleaning up after the job, so it probably is
not an issue for the results.

Other than that, the expansions/evaluations/costs are unchanged, coverage and
timeouts/memory-outs only changed by a couple of tasks, and I cannot see a
specific trend on the runtime.

http://ai.cs.unibas.ch/_tmp_files/pommeren/issue748-v1-opt-issue748-base-issue748-v1-compare.html
http://ai.cs.unibas.ch/_tmp_files/pommeren/issue748-v1-sat-issue748-base-issue748-v1-compare.html

http://ai.cs.unibas.ch/_tmp_files/pommeren/issue748-v1-sat-total_time-ehc-ff-issue748-base-issue748-v1.png

I think this can be merged.

Now that issue747 is merged I updated lazy search as well. I can now run
experiments on this. I'd suggest the same configurations as in issue747 for lazy
search; LM-cut and blind search on the optimal benchmarks for eager search; and
ehc(ff()) for EHC.

Code looks reasonable, probably wants an experiment.

I started with this already and created a pull request:
https://bitbucket.org/FlorianPommerening/downward-issues/pull-requests/39/issue748/diff

The pull request doesn't touch LazySearch yet. I'll do the same change as in the
other searches there once issue747 is merged.

This should be simple once we merged issue474. This is the last place where
global operators occur outside of the global task, so afterwards we could
continue moving the global task into its own class in issue700.