Investigate performance for high dimesions

Do only after issues #7, #15 (closed) have been implemented.

Possibly same as issue #2 (closed).

For the yTOR example with initially over 40 dimensions sampled and as few as nelip=5e3, or even nelip=5e2 for some projections, HYPERSPACE performance is bad, and eventually it looks like it gets stuck, in a CPU-intensive usage. That didn't happen with initially selected 25 out of 29 dimensions for model 2.

If it's not #2 (closed), it could be e.g. that one of the 4 (non-projectable) dimensions added to analysis is "difficult" (in terms of ellipsoids coverage), or some other MATLAB quirkiness for very long running processes. In these cases, preferably start with testing directly a high-dimensional "flat" problem with HYPERSPACE.