Make sampling kernels

[Teja10]

After computing ray-intersections, we need to sample points from the entry and exit points.
See the below for details on the API: https://docs.google.com/document/d/1SCwFPvVdABQrFdUJ3PKErpjh5RIekKGXvfDJECeT8PY/edit#heading=h.1gy1ce1y3h7w

[crozhon]

Wrote sampling kernels.

They expect start and end points as float3 and also require a thrust::minstd_rng to be created elsewhere and provided to the kernel. The launch is performed with the number of segments as the number of threads. (A segment being the line from a start point to an end point).

Output is a list of sample points, with NUM_SAMPLES_PER_SEGMENT * NUM_SEGMENTS (aka the number of threads) as the size.

For the volume rendering kernel (See #7), it's a lot easier to use the t-value of the ray rather than the point itself. So that kernel still assumes that all of the samples that it receives are numbers from 0.0 to 1.0, where 0.0 is the point where the ray enters the volume for the first time, and t=1.0 is the point where it leaves the volume. So at some point, we'll need to either generate those instead or convert the samples to those.