Weighted Lasso estimates for sparse logistic regression: non-asymptotic properties with measurement errors