GP covariance functions test cases

math

#1

This is directed toward @syclik from #1095, about writing test cases for the initial set of GP covariance functions. What would you like to see tested?

So far, I’ve got boundary conditions for invalid parameter values, a lot of tests for var types for autodiff, numerical accuracy. I’ve also gone through interfaces and tried invalid inputs, but I’m not sure how to write these in a gtest case.

What else would you like to see tested? What additions should I include in rev? Should I do any performance test in gtest?