DataSet.to_numpy() should use numpy dtypes whenever possible

[JBGreisman]

Pandas DataFrames that contain ExtensionDtypes always default to output data with object dtype when DataFrame.to_numpy() is called. This is suboptimal for MTZ data, which by construction must be compatible with float32, and possibly int32.

This PR wraps the pandas call with DataSet.to_numpy() to assess whether a more sensible default (either float32 or int32) can be used based on the existing data. This should help to avoid cases where data is unnecessarily cast to an object array, which can lead to unexpected behavior downstream.

[codecov-commenter]

Codecov Report

Merging #182 (7c0264f) into main (4222ffc) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #182      +/-   ##
==========================================
+ Coverage   98.36%   98.37%   +0.01%     
==========================================
  Files          45       45              
  Lines        1772     1783      +11     
==========================================
+ Hits         1743     1754      +11     
  Misses         29       29              

Flag	Coverage Δ
unittests	98.37% <100.00%> (+0.01%)	⬆️

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

Impacted Files	Coverage Δ
reciprocalspaceship/dataset.py	98.20% <100.00%> (+0.04%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[kmdalton]

I like it. I wish we also handled the case where MTZDtype is mixed with Pandas float dtypes. I'm not sure that can be implemented in a way that doesn't cause nasty side effects.