Differentially Private $\ell_1$-norm Linear Regression with Heavy-tailed Data