Slit resolution calculation fails for q_length=0 and small q_width.

[pkienzle]

The following comment will be added to resolution.py (#533):

elif li == 0:
    in_x = 1.0 * ((q_calc >= qi-wi) & (q_calc <= qi+wi))
    abs_x = 1.0*(q_calc < abs(qi - wi)) if qi < wi else 0.
    # TODO: need pixel fraction for the boundary pixels
    # Diff q_edges in the expression below gives the pixel width in
    # q. As wi drops below the width of a single pixel, the
    # numerator stays the same but the denominater gets smaller, leading
    # to ever increasing weights. We can't see this in SasView since
    # it explicitly excludes q_length < 10 q_width.
    weights[i, :] = (in_x + abs_x) * np.diff(q_edges) / (2*wi)

We should either fix the code or disallow this configuration of slits. As noted, SasView forces q_length > 10 q_width.

For q_width large relative to the q spacing the issue is not significant.

The same issue doesn't appear for q_width=0 and q_length small even though it uses a similar diff(u_edges) / length construction.