clip I for Sigma calc

[DHekstra]

When using scale_merged_intensities to perform French-Wilson scaling in the anisotropic mode, the local Sigma is calculated by a local weighted average of observed intensities. As a consequence, Sigma can be negative. This leads to NaNs in the calculation of the logpdf used to calculate the posterior expected intensity. To remedy this, I propose clipping the observed intensities to be positive in (and only in) the calculation of Sigma in mean_intensity_by_miller_index(I, H, bandwidth).

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (0e2f5e2) 92.39% compared to head (d25d7da) 92.40%.
Report is 3 commits behind head on main.

❗ Current head d25d7da differs from pull request most recent head f7f472e. Consider uploading reports for the commit f7f472e to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #235   +/-   ##
=======================================
  Coverage   92.39%   92.40%           
=======================================
  Files          37       37           
  Lines        2434     2435    +1     
=======================================
+ Hits         2249     2250    +1     
  Misses        185      185           

Flag	Coverage Δ
unittests	92.40% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kmdalton]

I'm not certain that fw should really be applied to data with resolution shells that have negative average intensity. Silently clipping the values is not a good solution if it leads to the user unknowingly processing such data. If you want to merge a change like this, it needs to print a warning or better yet be implemented as an optional flag which is disabled by default.

[DHekstra]

That makes sense. In my single test, Sigma was negative for ~0.01% of reflections. I'll think some more about this when time permits.