19 remove orphaned cov argument from normallhs sampling

[ClaasRostock]

• removed orphaned _cov argument from normalLhs sampling
• updated to dictIO>=0.3.0 (from dictIO>=0.2.9)

[frl000]

Ah, good point. I forgot that. It is okay for me If we drop that cov argument. If there is one day a need to get it back, we will find.
The basic idea was besides scaling and stretching also to apply rotation by simple sampling setup. Otherwise one can achieve that with latter matrix ops as well.
The reason why I did not implement that from the beginning was the limitation of
scipy.stats.norm(loc=mean, scale=std).ppf(distribution),
only taking (nx1) vectors for mean and scale. Rotation would require a (nxn) matrix.
I write this here as a reminder...