This release modernizes the deeplabcut-docker helper. The vendored shell script

(deeplabcut_docker.sh) has been replaced with a pure-Python implementation,

and the package setup has been updated to use pyproject.toml.

- The package now requires Python ≥ 3.10.

- `setup.cfg` / `MANIFEST.in` have been replaced by `pyproject.toml`.

Re-install from the new package to pick up the updated entry point.

- **Pure-Python rewrite** — replaced the vendored shell script with a single

`deeplabcut_docker.py` module; no shell dependency required.

- **`--image` flag** — pass an arbitrary `repo:tag` or digest to override the

default image selection from `DLC_VERSION` / `CUDA_VERSION`.

- **Local image support** — skips the registry pull when a matching image is

already present locally (or when a specific version/image is requested).

- **Supplementary group forwarding** — all supplementary groups of the current

host user are forwarded into the container via `--group-add`.

- **Jupyter image validation** — warns at startup when the pulled image does

not appear to have a Jupyter entrypoint.

- **Configurable Docker binary** — override via `DOCKER` environment variable

(e.g. `DOCKER="sudo docker"`).

- **Token privacy** — the notebook token is no longer echoed to the terminal by

default; users are instructed to use the value of `NOTEBOOK_TOKEN` instead.

- **Empty-token warning** — setting `NOTEBOOK_TOKEN=` (empty string) disables

Jupyter token authentication. `deeplabcut-docker notebook` now logs a

prominent warning in this case.

- Fixed broken Jupyter image support on the current Docker Hub tags.

- Fixed default user home directory (`/home/{user}` instead of root).

- Fixed notebook token not being propagated into the container environment.

- Improved error logging and exit-status propagation from `docker` subprocesses.

- Replaced `setup.cfg` + `MANIFEST.in` with `pyproject.toml`.

- Added `LICENSE` (LGPL-3.0) and `Makefile` for local build and PyPI release.