Refine pyproject.toml Ruff config and optional dependencies (fixes #542)

[sannithshetty]

This PR addresses the feedback from Issue #542 (“Technical and Configuration Feedback on pyproject.toml”).

Changes

1. Ruff configuration

   • Removed the [tool.ruff.lint] sub-table.
   • Moved all lint options into [tool.ruff] using the lint.* key style (e.g. lint.select, lint.ignore).
   • Ensured dummy-variable-rgx is defined under [tool.ruff] so it is correctly picked up by Ruff.

2. Platform-specific dependencies

   • Grouped platform-specific packages into project.optional-dependencies:
     • ros → cv_bridge, eclipse-zenoh
     • hardware → hid, py-mjpeg
     • jetson → tf_keras (and related Jetson-only deps, if applicable)
   • Kept the base [project].dependencies list focused on cross-platform packages.

3. Deep learning libs & compatibility

   • Reviewed torch, torchvision, tensorflow usage.
   • Did not hard-pin versions in this PR to avoid breaking existing setups, but left comments in pyproject.toml indicating that version constraints should follow the officially tested matrix in the docs.

Rationale

• Ruff prefers a flat [tool.ruff] configuration for reliability and aligns better with the current docs.
• Moving ROS / hardware / Jetson dependencies into extras prevents install failures on platforms where these packages are unavailable.
• This keeps the default uv sync flow smooth while still allowing advanced users to opt into platform-specific extras.

Testing

• uv sync
• uv run pre-commit run --all-files
• uv run pytest (if applicable on my environment)

Fixes #542.