Implemented yolo dataset support

[cansik]

I have implemented yolo dataset support to load data in the YOLO format as mentioned in #10. I have used the existing XML-to-COCO dataset and adapted it to support the yolo format.

The loader expects the image files to be either ".png", ".jpg", ".jpeg", ".bmp", ".tiff" and tries to find the best-matching one. Usually the img_path and ann_path are the same and contain .txt files as well as the image files.

Loading YOLO datasets can be slower because we have to open every image to determine the image size. This could be enhanced by using multi-processing, which currently is not implemented.

[RangiLyu]

Thanks for your contribution! Would you like to add a brief introduction to the Yolo format dataset in the README?
You can add the introduction at here:

nanodet/README.md

Line 218 in d8ba391

1. **Prepare dataset**

Also, please run pip install pre-commit and then run pre-commit run --all-files to fix the lint.

[cansik]

Thank you for the hint about the formatting, I have updated the code using the pre-commit command.
Btw it was necessary to update the isort package to 5.12.0: PyCQA/isort#2077

[cansik]

@RangiLyu Readme has been adjusted as well. Would be great if you could have a look again.

[codecov]

Codecov Report

Merging #487 (2357abd) into main (d8ba391) will decrease coverage by 1.05%.
The diff coverage is 20.87%.

@@            Coverage Diff             @@
##             main     #487      +/-   ##
==========================================
- Coverage   74.65%   73.60%   -1.05%     
==========================================
  Files          70       71       +1     
  Lines        4600     4691      +91     
  Branches      716      728      +12     
==========================================
+ Hits         3434     3453      +19     
- Misses        975     1045      +70     
- Partials      191      193       +2     

Flag	Coverage Δ
unittests	73.60% <20.87%> (-1.05%)	⬇️

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

Impacted Files	Coverage Δ
nanodet/data/dataset/__init__.py	47.82% <20.00%> (-7.73%)	⬇️
nanodet/data/dataset/yolo.py	20.93% <20.93%> (ø)
nanodet/util/logger.py	66.41% <0.00%> (-0.25%)	⬇️
nanodet/trainer/task.py	55.23% <0.00%> (ø)
nanodet/util/config.py	84.00% <0.00%> (+0.66%)	⬆️

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