feat: Added pretrained PyTorch mobilenets

[fg-mindee]

Following up on #255, this PR introduces the following modifications:

• adds pretrained params for PyTorch mobilenets (character classification)
• updated unittests

import torch
from doctr.models import mobilenet_v3_small
from doctr.datasets import VOCABS
from doctr.datasets.classification.base import synthesize_char_img
from doctr.io import tensor_from_pil
from torchvision.transforms import Normalize

model = mobilenet_v3_small(pretrained=True).eval()
img = synthesize_char_img('a', font_family='FreeMono.ttf')
transfo = Normalize(mean=(0.694, 0.695, 0.693), std=(0.299, 0.296, 0.301))

with torch.no_grad():
    out = model(transfo(tensor_from_pil(img)))

print(VOCABS['french'][out.argmax().item()])

yields:

'a'

Any feedback is welcome!

[codecov]

Codecov Report

Merging #415 (fe64d05) into main (350a961) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #415      +/-   ##
==========================================
+ Coverage   95.78%   95.83%   +0.04%     
==========================================
  Files          95       96       +1     
  Lines        3917     3937      +20     
==========================================
+ Hits         3752     3773      +21     
+ Misses        165      164       -1     

Flag	Coverage Δ
unittests	95.83% <100.00%> (+0.04%)	⬆️

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

Impacted Files	Coverage Δ
doctr/models/backbones/mobilenet/__init__.py	100.00% <100.00%> (ø)
doctr/models/backbones/mobilenet/pytorch.py	100.00% <100.00%> (ø)
doctr/models/core.py	94.95% <0.00%> (+0.84%)	⬆️

---

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 350a961...fe64d05. Read the comment docs.