A Geometric Approach to Pairwise Bayesian Alignment of Functional Data Using Importance Sampling