varianceSimulation: Variance simulation
In schwenzfeier/udp: Statistically Useful Differential Privacy

Estimates the variance of \tilde θ^{dp} via simulation.

1 2	varianceSimulation(theta_tilde, sigma_hat, theta_hat_dp_overall, a_1, a_2, P, lambda, theta_noise, alpha_noise, nsims)

`theta_tilde`	Estimate of QOI
`sigma_hat`	Estimate of variance
`theta_hat_dp_overall`	Differentially private censored QOI
`a_1`	DP estimate of proportion of partitions that were left-censored
`a_2`	DP estimate of proportion of partitions that were right-censored
`P`	Number of partitions
`lambda`	Bounding parameter
`theta_noise`	SD of dp noise added to theta_hat_dp
`alpha_noise`	SD of dp noise added to alpha_dp
`nsims`	Number of simulations to boostrap variance

`var_est`	Variance estimate of theta_tilde
`theta_tilde_sims`	Vector of simulated theta_tilde estimates
`mvn_draws`	Matrix of theta_hat_dp and alpha draws from the multivariate normal
`mat_not_pos_def`	Indicator for whether an adjustment was applied to fix numerical issue where the covariance matrix can appear to be not positive definite
`sigma_matrix`	Covariance matrix used in the multivariate normal draws
`var_theta_hat_nonoise`	Estimated variance of theta_hat_dp not including variance from dp noise added
`orig_sigma_matrix`	Covariance matrix prior to adjustment to be positive definite