refactor: Refactored API of doctr.models

[fg-mindee]

This PR introduces the following modifications:

• cleaned the __init__ files of doctr.models to keep only the relevant modules accessible (private modules that will see their API change in the future are better left out of this)
• moved DocumentBuilder from doctr/models/core.py to doctr/models/builder.py
• merged DetectionModel and RecognitionModel into BaseModel
• each predictor class is now split into its implementations in PyTorch & TensorFlow via the same dispatch mechanism as previously (for more specific operations, typing and interface)
• move crop split from doctr/models/recognition/core.py to doctr/models/recognition/base.py
• updated unittests
• updated API
• updated demo

Any feedback is welcome!

[codecov]

Codecov Report

Merging #481 (1bcbbe4) into main (e134da4) will decrease coverage by 0.44%.
The diff coverage is 83.97%.

@@            Coverage Diff             @@
##             main     #481      +/-   ##
==========================================
- Coverage   95.42%   94.97%   -0.45%     
==========================================
  Files          97      108      +11     
  Lines        4065     4160      +95     
==========================================
+ Hits         3879     3951      +72     
- Misses        186      209      +23     

Flag	Coverage Δ
unittests	94.97% <83.97%> (-0.45%)	⬇️

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

Impacted Files	Coverage Δ
doctr/models/__init__.py	100.00% <ø> (ø)
doctr/models/detection/__init__.py	100.00% <ø> (ø)
doctr/models/recognition/__init__.py	100.00% <ø> (ø)
doctr/models/predictor/pytorch.py	38.46% <38.46%> (ø)
doctr/models/recognition/predictor/base.py	51.61% <51.61%> (ø)
doctr/models/recognition/predictor/pytorch.py	87.87% <87.87%> (ø)
doctr/models/recognition/predictor/tensorflow.py	91.17% <91.17%> (ø)
doctr/models/builder.py	93.00% <93.00%> (ø)
doctr/models/detection/predictor/pytorch.py	94.44% <94.44%> (ø)
doctr/models/core.py	100.00% <100.00%> (+5.38%)	⬆️
... and 24 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 e134da4...1bcbbe4. Read the comment docs.

[charlesmindee]

thanks for the PR!

[charlesmindee]

Thanks!