Added quasi-random number generator and support for Rand<double>.

[paulhoux]

Based on the excellent article by Martin Roberts: http://extremelearning.com.au/unreasonable-effectiveness-of-quasirandom-sequences/

The QuasiRand class returns quasi random values in the range [0, 1) which are evenly distributed across the full range. When used to generate points on a sphere, it looks like this:

By comparison, fully random points would look like this:

The great thing about the algorithm is that it doesn't require you to know the number of values up front. Here's the same sphere after simply adding 4 times as many points:

The class is templated and uses float by default, but also supports double and long double.

I will add a sample later.

Also refactored the Rand class by adding a template argument. This allows the use of double and long double random values. By default it uses float values.

[richardeakin]

Really cool. And I bet you could use something like a combination of QuasiRand + Rand to create a sort of blue noise.

One thought - does it need to live in a separate file? Doesn't seem like much code, and I would think this, along with similar random number generators, could all live within cinder/Rand.h / cinder/Rand.cpp

[paulhoux]

Actually I believe this quasi-random stuff should remain in its own .h and .cpp files. It's different enough from Rand and should be a deliberate opt-in, in my opinion. I guess it's a matter of taste.

[totalgee]

Just a side note -- I've used the very good PCG random generator (http://www.pcg-random.org or https://github.com/imneme/pcg-cpp) in Cinder projects, when I needed random generation that was fast and produced the same results (for the same seeding) on different platforms (Mac and Windows, in my case). You don't get that with the standard library (for example std::mt19937, which I believe is what ci::Rand uses). It's easy enough to use (the header) on its own, but if you're "revisiting" or adding random stuff, it might be something to think about for Cinder...e.g. as a drop-in replacement for Rand?

That said, the Martin Roberts R₂ "golden ratio" noise thing is also a nice addition, thanks! (-;