FIX: Drop header before resampling image to avoid unsafe cast

[effigies]

int16 images with scale factors can get badly truncated or wrapped. This PR scales the data array and drops the header before passing to Affine().resample() to avoid this. We then re-add the dtype, though this may not be necessary. I think the effects of multiple interpolation are going to be minimal here, so keeping dtypes as expected is probably worth it.

[oesteban]

Is this fixed in nitransforms?

[codecov]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (5ef205f) 76.62% compared to head (efa3528) 76.59%.

Files	Patch %	Lines
sdcflows/interfaces/utils.py	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #415      +/-   ##
==========================================
- Coverage   76.62%   76.59%   -0.03%     
==========================================
  Files          32       32              
  Lines        2823     2824       +1     
  Branches      373      373              
==========================================
  Hits         2163     2163              
- Misses        594      595       +1     
  Partials       66       66              

Flag	Coverage Δ
unittests	?

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.

[effigies]

Not yet. Since that needs to work more broadly, a deeper review will be necessary.