fix: Fixed edge case for detection

[fg-mindee]

This PR introduces the following modifications:

• fixes the crop extraction in OCRPredictor: the test is not always failing because this part is not executed if it didn't find any box. Fixed it by checking whether there is some, and if so removing the objectness before passing it
• fixes the coord condition in unittesting: rarely but they can take the values 0 & 1, so the strict inequalities were changed to looser conditions.

Any feedback is welcome!

[codecov]

Codecov Report

Merging #44 (db92989) into main (3868d9e) will increase coverage by 0.19%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #44      +/-   ##
==========================================
+ Coverage   97.75%   97.94%   +0.19%     
==========================================
  Files          20       20              
  Lines         534      536       +2     
==========================================
+ Hits          522      525       +3     
+ Misses         12       11       -1     

Impacted Files	Coverage Δ
doctr/models/core.py	97.22% <100.00%> (+2.77%)	⬆️
doctr/models/detection/core.py	93.33% <100.00%> (ø)
...tr/models/detection/differentiable_binarization.py	98.18% <100.00%> (ø)
doctr/models/recognition/core.py	92.59% <100.00%> (ø)
doctr/models/recognition/crnn.py	100.00% <100.00%> (ø)
doctr/models/recognition/sar.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5e467b7...db92989. Read the comment docs.