feat: save logo text (extracted using OCR in DB)

[raphael0202]

Save in DB in a new field logo_annotation.text the text extracted using OCR.
It's useful as we can have a much higher confidence in the logo prediction with the text included.

I also added integration to a new repository test-data, that is going to replace the use of git LFS. Unit/ML/integration tests will fetch assets from this repository.
The assets are cached locally using diskcache, so that they don't have to be downloaded again.

[codecov]

Codecov Report

Merging #1283 (c5923b6) into main (e4080ca) will increase coverage by 0.04%.
The diff coverage is 89.28%.

❗ Current head c5923b6 differs from pull request most recent head 389aaaa. Consider uploading reports for the commit 389aaaa to get more accurate results

@@            Coverage Diff             @@
##             main    #1283      +/-   ##
==========================================
+ Coverage   56.61%   56.66%   +0.04%     
==========================================
  Files          77       78       +1     
  Lines        7547     7566      +19     
==========================================
+ Hits         4273     4287      +14     
- Misses       3274     3279       +5     

Files	Coverage Δ
robotoff/models.py	95.51% <100.00%> (+0.02%)	⬆️
robotoff/prediction/object_detection/core.py	50.00% <100.00%> (ø)
robotoff/utils/__init__.py	61.19% <100.00%> (ø)
robotoff/utils/cache.py	100.00% <100.00%> (ø)
robotoff/utils/download.py	100.00% <100.00%> (ø)
robotoff/utils/image.py	79.41% <80.00%> (-9.30%)	⬇️
robotoff/workers/tasks/import_image.py	30.58% <54.54%> (+0.99%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more