Pre-release fixes & improvements checklist

[C-Achard]

• Update & check pyproject.toml (specify tf/pytorch installation)
• Check automated torch device discovery logic -> in DLC-live
• Improve the selection of which camera is used for inference (e.g. dropdown menu)
• TCP sockets clearing & proper handling
• Properly handle releasing writer when inference is interrupted
• Fix/disentangle FPS counters of multiple cameras
• Setting custom FPS handling when unsupported (avoid setting wrong values)
• On OpenCV backend, exposure time and gain should not be configurable
• Fix OpenCV logging deprecation
• Switch to typed configs
• Make it even prettier
• Saving configs, video, timestamps, ... in a standard folder
  • Add toggle to add timestamp to filename