probe.oversample should use the resolution type when generating calculation points

[pkienzle]

We are not modifying the set of Q points used for the convolution when the form of the resolution function is changed. We should be ensuring that we have enough calculated points to support the ΔQ at each measured point. For a normal distribution this requires a wider window with points concentrated at the center. For a uniform distribution they should be sampled uniformly within the resolution width. The method probe.oversample assumes a normal distribution.