Retain image_ID from input file

[sfmig]

When loading an annotation file into a df/ds, optionally allow users to keep the image_ID as in the input file (right now, the image ID is reassigned based on the list of sorted image filenames).

One easy way would be to compute the mapping between filename and imageID at the start and apply it to the df/ds at the end if required

[GauravSRC]

Hi @sfmig and maintainers - I implemented an opt-in retain_image_id feature for bounding box loaders and opened a PR:#138

Summary:
from_files(..., retain_image_id=True) preserves COCO images[].id values (and numeric VIA keys where possible).
Default behaviour is unchanged (retain_image_id=False).
Added a unit test tests/test_retain_image_id.py.
Ran pytest locally (170 tests passed) and fixed pre-commit formatting issues locally. I also removed accidentally tracked .venv files so packaging checks pass.
This is a minimal, backwards-compatible change; happy to adjust the behaviour if you prefer a different mapping approach (e.g., also returning a map from original IDs --> ethology indices). Please let me know if you'd like any changes or if I should add documentation/examples in the repo docs.