A researcher is studying the effect of a new fertilizer on plant height. The model is $Y_i = \beta_0 + \beta_1 X_i + \epsilon_i$ where $X_i$ is fertilizer amount and $Var(\epsilon_i) = \sigma^2 X_i^2$. Why is the OLS estimator $\hat{\beta}$ no longer the Best Linear Unbiased Estimator (BLUE)?