Fix geometric extrapolation when points-per-decade is provided or data is a single point

[pkienzle]

I noticed that the log_delta_q is calculated as the inverse of the equation listed in the documentation:

$$ \log \Delta q = (\log q_n - \log q_1) / (n - 1) $$

This works out as long as this condition for calculating log_delta_q is met as the calculations for n_low and n_high also use the inverse term appropriately. However, if this condition is not met, then log_delta_q is calculated by:
log_delta_q = log(10.) / DEFAULT_POINTS_PER_DECADE
or
log_delta_q = log(10.) / points_per_decade
which are no longer the inverse term. This would result in an incorrect calculation of n_low and n_high.

For example:

In[2]: import numpy as np
In[3]: q = np.logspace(-5, -4, 11)
In[4]: import sasmodels.resolution as resolution
In[5]: resolution.geometric_extrapolation(q, 1e-6, 1e-3)
Out[5]: 
array([1.00000000e-06, 1.25892541e-06, 1.58489319e-06, 1.99526231e-06,
       2.51188643e-06, 3.16227766e-06, 3.98107171e-06, 5.01187234e-06,
       6.30957344e-06, 7.94328235e-06, 1.00000000e-05, 1.25892541e-05,
       1.58489319e-05, 1.99526231e-05, 2.51188643e-05, 3.16227766e-05,
       3.98107171e-05, 5.01187234e-05, 6.30957344e-05, 7.94328235e-05,
       1.00000000e-04, 1.25892541e-04, 1.58489319e-04, 1.99526231e-04,
       2.51188643e-04, 3.16227766e-04, 3.98107171e-04, 5.01187234e-04,
       6.30957344e-04, 7.94328235e-04, 1.00000000e-03])
In[6]: resolution.geometric_extrapolation(q, 1e-6, 1e-3, points_per_decade=10)
Out[6]: 
array([1.00000000e-06, 1.00000000e-05, 1.25892541e-05, 1.58489319e-05,
       1.99526231e-05, 2.51188643e-05, 3.16227766e-05, 3.98107171e-05,
       5.01187234e-05, 6.30957344e-05, 7.94328235e-05, 1.00000000e-04,
       1.00000000e-03])

@pkienzle can you confirm I am understanding intended functionality correctly? I think this issue would pre-date this PR.

Originally posted by @caitwolf in #563 (comment)