update `download` method to support nested file structure

[csbrown]

When I create a COCO-format file for my images, if the filenames are nested, then the download method fails.

For example, with the following COCO-format file named 'tiny_image.json':

{
  "info": {},
  "images": [
    {
      "id": 1,
      "file_name": "yummy/taco.jpg",
      "coco_url": "https://en.m.wikipedia.org/wiki/File:NCI_Visuals_Food_Taco.jpg"
    }
  ],
  "annotations": [],
  "licenses": []
}

Note the file_name, which contains a nested file.

When attempting to use the download method, we get the following:

>>> from pycocotools.coco import COCO
>>> coco = COCO('tiny_image.json')
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
>>> coco.download('.')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/scott/NOAA/research/deep7_modeling/venv/lib/python3.11/site-packages/pycocotools/coco.py", line 390, in download
    urlretrieve(img['coco_url'], fname)
  File "/usr/lib/python3.11/urllib/request.py", line 251, in urlretrieve
    tfp = open(filename, 'wb')
          ^^^^^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: './yummy/taco.jpg'

This is because the urlretrieve method does not automatically create the necessary directory structure. I think a one or two line fix of:

os.makedirs(os.path.dirname(filename), exist_ok=True)

On the line immediately before the call to urlretrieve will enable this. Happy to submit a PR if there is a maintainer willing to pull it.