Fix threshold computation handling nan values

[giodavoli]

Closes #327

Replaced np.std and np.mean respectively with np.nanstd() and np.nanmean()

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (dd20ef7) 83.40% compared to head (1254121) 82.55%.
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #333      +/-   ##
==========================================
- Coverage   83.40%   82.55%   -0.86%     
==========================================
  Files         100      100              
  Lines        7245     7550     +305     
  Branches     1275     1353      +78     
==========================================
+ Hits         6043     6233     +190     
- Misses        905      980      +75     
- Partials      297      337      +40     

Files	Coverage Δ
nannyml/thresholds.py	85.71% <100.00%> (ø)

... and 4 files with indirect coverage changes

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

[nnansters]

Hey @giodavoli , just added a quick test to numerically verify the fix. You raised an excellent point on about dealing with NaN's, it's something we've not tested enough for. I'll add this to the engineering backlog. In the meantime if you run into issues, don't hesitate to log and maybe even propose a fix.

In any case: thank you very much for taking the time to investigate, report and eventually fix this issue!