Better Benchmarking

I wanted to make the speed benchmarking more automated. Here's how it works: 1. On the master branch, run `make set-bench`. This will run a multivariate Gaussian several times, and create a (gitignored) file saving info about the run and your machine. Because it's gitignored, it will just stay on your machine, even when you change branches. 2. Once you've made a change, on a new branch, run `make compare-bench`. This will run the same Gaussian, and if it's on average more than 2% slower than the existing benchmark, it will fail. For example, I might attempt some refactoring of the code in the future, and if I do, I will follow the above. That way, I can be sure that the code is not significantly slower.