Add bootstrap performance functionality

[mccarthy-m-g]

For model comparisons (#26), we'll need to be able to calculate confidence intervals for various error metrics (e.g., rmse). We have some internal tooling for this we can move over and adapt for mipdeval