How the augmented Lagrangian algorithm can deal with an infeasible convex quadratic optimization problem — Motivation, analysis, implementation