Seeding improvements

[claresinger]

1. Shuffle logic: e.g. don't need to shuffle if only one type of seed particle or if the number of super particles to inject is equal to the number of possible seeds.
2. Example comparison from literature (maybe https://doi.org/10.1098/rsta.2012.0086), some other candidate papers listed in Time-dependent super-droplet sampling and modification #1331
3. Example of seeding in 1D and 2D domains
4. Parallelisation of the sampling logic
5. GPU support

[jtbuch]

I've made a PR addressing point 3. above in #1396

[slayoo]

@jtbuch, @claresinger, one paper that, IIRC, we have not mentioned yet discussing seeding in PySDM is the Andrejczuk et al. 2014 super-droplet seeding/geoengineering study: https://doi.org/10.1016/j.atmosres.2014.01.006

[jtbuch]

Thanks for the reference, @slayoo! I will incorporate seeding in the 2D case and try to reproduce their results.