Fix broken link of Wainwright's paper

examples/linear_model/plot_sparse_recovery.py

@@ -15,7 +15,7 @@
 number of samples, the number of features, the conditioning of the
 design matrix on the signal subspace, the amount of noise, and the
 absolute value of the smallest non-zero coefficient [Wainwright2006]
-(http:https://statistics.berkeley.edu/tech-reports/709.pdf).
+(http:https://statistics.berkeley.edu/sites/default/files/tech-reports/709.pdf).
 
 Here we keep all parameters constant and vary the conditioning of the
 design matrix. For a well-conditioned design matrix (small mutual