Account for bbox when saving predictions

[AdeelH]

Overview

When predictions are made on a cropped RasterSource (i.e. with a bbox specified), that bbox must be accounted for when saving those predictions; otherwise, the georeferencing will be wrong.

Checklist

• Added needs-backport label if PR is bug fix that applies to previous minor release
• Ran scripts/format_code and committed any changes
• Documentation updated if needed
• PR has a name that won't get you publicly shamed for vagueness

Notes

N/A

Testing Instructions

• These changes were manually tested. Without these changes the prediction outputs were misaligned with the source imagery. With these changes, they were correctly aligned.

[codecov]

Codecov Report

Merging #1931 (90f2487) into master (1d23e46) will decrease coverage by 0.01%.
The diff coverage is 77.77%.

@@            Coverage Diff             @@
##           master    #1931      +/-   ##
==========================================
- Coverage   82.47%   82.47%   -0.01%     
==========================================
  Files         190      190              
  Lines        9325     9330       +5     
==========================================
+ Hits         7691     7695       +4     
- Misses       1634     1635       +1     

Files	Coverage Δ
...sion/core/data/label/chip_classification_labels.py	100.00% <100.00%> (ø)
...rvision/core/data/label/object_detection_labels.py	97.12% <100.00%> (ø)
...on/core/data/label/semantic_segmentation_labels.py	91.53% <ø> (ø)
...a/label_store/chip_classification_geojson_store.py	100.00% <ø> (ø)
...data/label_store/object_detection_geojson_store.py	97.14% <ø> (ø)
...n_core/rastervision/core/data/label_store/utils.py	90.00% <100.00%> (ø)
...a/label_store/semantic_segmentation_label_store.py	91.62% <66.66%> (-0.91%)	⬇️

... and 1 file with indirect coverage changes