Lambertian BRDF is actually cosine squared brdf

[vchizhov]

The Lambertian brdf uses random_in_unit_sphere, however this induces a cosine squared brdf rather than a Lambertian one. To correct this, one has to use random_on_unit_sphere (this can be achieved through normalize(random_in_unit_sphere)). For more information see: https://github.com/vchizhov/Derivations/blob/master/Probability%20density%20functions%20of%20the%20projected%20offset%20disk%2C%20circle%2C%20ball%2C%20and%20sphere.pdf
Note that this is a mistake in both the book and the code.

[trevordblack]

We are aware of the problem.
This (#145). #155, and #179 are all probably going to make the next release (2.1.0 or 2.0.1).
We are in no position to merge anything for the next few weeks.

[hollasch]

Resolved in PR #229, development branch.