refactor: Drops support of np.float16

[fg-mindee]

This PR introduces the following modifications:

• removed fp16 support from datasets and models' preprocessors
• removed support of np.float16 for targets in text detection models
• adapted unittests

Any feedback is welcome!

[codecov]

Codecov Report

Merging #627 (af1f665) into main (f8f6d3e) will decrease coverage by 0.09%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #627      +/-   ##
==========================================
- Coverage   96.19%   96.09%   -0.10%     
==========================================
  Files         112      112              
  Lines        4335     4328       -7     
==========================================
- Hits         4170     4159      -11     
- Misses        165      169       +4     

Flag	Coverage Δ
unittests	96.09% <100.00%> (-0.10%)	⬇️

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

Impacted Files	Coverage Δ
doctr/datasets/cord.py	97.43% <ø> (-0.07%)	⬇️
doctr/datasets/datasets/base.py	95.34% <100.00%> (-0.11%)	⬇️
doctr/datasets/datasets/pytorch.py	100.00% <100.00%> (ø)
doctr/datasets/datasets/tensorflow.py	100.00% <100.00%> (ø)
doctr/datasets/detection.py	96.87% <100.00%> (-0.10%)	⬇️
doctr/datasets/doc_artefacts.py	94.11% <100.00%> (ø)
doctr/datasets/iiit5k.py	96.77% <100.00%> (ø)
doctr/datasets/ocr.py	88.00% <100.00%> (ø)
doctr/datasets/recognition.py	76.00% <100.00%> (-16.00%)	⬇️
doctr/datasets/sroie.py	94.87% <100.00%> (ø)
... and 6 more

---

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 f8f6d3e...af1f665. Read the comment docs.

[charlesmindee]

Thanks!

[fg-mindee]

cf #263