Christ, finally

All the benchmarks are running again.

Adding NUMA support took SOOO long, OMG I don’t even etc.

(So the benchmarks can now run SMP, NUMA or NUMA unused, where the benchmark admin data is NUMAed but the data which is actually liblfds data is only one the NUMA node which has the largest number of logical cores in the current thread set).

Need to fix up the gnuplots now, add a benchmark duration argument, some other tidying up.

Then I can make the docs, run the tests, then publish 7.1.0 – and then I can have snazzy glittery gnuplots on the home page!