Pecuzal has samplesize keyword

[hkraemer]

In order to speed up computations for PECUZAL I implemented the samplesize-keyword, too. This is affecting the pecora()-call, which is fine since #103 but also the function uzal_cost_pecuzal. Here I found out that the NN-search was carried out for each time horizon T, which is absolutely not necessary. So I dragged this search out of the loop. Turns out that PECUZAL is now way faster, up to a factor of 2. Pretty stupid that I haven't double checked this on the inital implementation. Tests have been adjusted also and I chose to make only a minor version increment, since this is more a bugfix. But actually there is a new keyword, too, so you decide.

[Datseris]

Let's increment minor version instead for this one. Can you please write 1.0 instead of 1.? (and in general for all floats end in .0 instead of .)