Fix issue #138

This MR fixes #138 (closed). Also I added yade --quickperformance -j8 test to the pipeline. So that we will be sure that situation like with #138 (closed) and 002c1c94 will not happen again.

The first push is without the fix, only to confirm that the pipeline fails without the fix. Next commit will be with a fix.

I discovered this by accident because want to make a meaningful benchmark of high precision Real branch. I expect that vectorization can bring us some serious speed improvements. So I added an option yade --stdperformance it will keep retrying to run the benchmark until standard deviation of the performance is below 2.0%.

Also I moved make doc to be the first one in ASAN, so that we will see more meaningful output at the end of tests.