Stable convergence of generalized stochastic integrals and the principle of conditioning: L 2 theory