Added benchmark

[ghost]

The benchmark is used for testing BVH build and render times for various algorithms.

It downloads a Sponza model at configuration time, as opposed to storing it in the repository. That way, cloning the project doesn't take forever. Users who won't be running the benchmark won't want to wait for the long download time unless their running the benchmark.

The benchmark itself is made for comparing multiple build algorithms. It times the construction of the BVH as well as rendering time. The scheduler for the benchmark is multi-threaded, so it runs much faster than the scheduler in the examples. It also has progress updates for the tracing, which is nice for longer render times.

Initially, I had planned on running the "dummy" BVH build algorithm for a frame of reference. I ended up commenting it out because it was taking too long to render. I may still use the "dummy" build algorithm for integration testing later on, since it always produces a "known good" image.

Example output for a 100x100 pixel test:

Running benchmark for strategy 0
  Percent Completion: %100.00
Running benchmark for strategy 1
  Percent Completion: %100.00
Results:
              | Build Time (sec) | Render Time (sec) |
              |------------------|-------------------|
  Strategy 00 |   0.0080         |  12.3891          |
  Strategy 01 |   0.0725         |   0.0104          |

This PR addresses the second item in #14

[ghost]

Closing this, since I read #18 a little bit too late! Holding out for the bench marking system described there, as it sounds like it will be much more thought out. Sorry about that!

Moved to https://github.com/tay10r/Fast-BVH-benchmark

[brandonpelfrey]

I think this is a great starting point for #18, and we can use it as a jumping off point for more of what was described there. If you don't see any concerns, let's pull from your -benchmark repo. Sound good?

[ghost]

@brandonpelfrey If you insist! I'll reopen the PR.