Refactor debrief: known issues and potential improvements

[C-Achard]

This is meant to help list potential future additions and fixes to be made at a glance.

Potential features

• Skeleton display (see Seb/display skeleton #92 and Add option to visualize skeletons (edges connecting keypoints) #5)
• Undo (CTRL+Z)
• "Project manager" GUI, showing project files at a glance and providing quick actions
• Clustering (see clustering in napari-DLC GUI #43 and clustering feature in napari deeplabcut #38)
• Frame removal utility (remove unwanted frames for extraction)
• Some built-in evaluation tools (on machinelabels)

Known issues

Bugs

• The color scheme list highlight color does not match actual kp colors when loading a default config Update ClickableLabel in color scheme reference: don't use colored text #206
• Stray Qt pop-ups when opening, likely due to layout addition ordering. Might also create delay in opening
• Video frame extraction : if no config.yaml is present, extracting frames should ask for a config and save relative to it ideally
• Bodypart/individual toggle menu is sometimes incorrectly shown on start
• If KeypointControls were added by drag-and-drop, one can still open a new widget instance
• Moving a keypoint on frame t, then going to t+1 in LOOP mode, does not add a keypoint and skips to next frame immediately.

For future PRs

• _widgets.py is still too heavy and would benefit from further separation into other files
  • Started in Plugin refactor [8]: Improve layer management/lifecycle #186
• Annoying warning napari\_qt\widgets\qt_progress_bar.py:55: UserWarning: constrained_layout not applied because axes sizes collapsed to zero. Try making figure larger or Axes decorations smaller.! should be fixed (or silenced, failing that).
• While documentation warns not to load several datasets at once, there is no hard guard either; this should be made safe or enforced by the code itself.
• More ergonomic labeling controls (e.g. using A/D to navigate frames, W/S to switch keypoints, and Shift/Space for special actions or saving) could be nice
• Machine labels should have visually different markers
• Disable/monkeypatch "5" keybind shortcut for transform tool ? -> Show warning
• Centralizing logic further would be valuable. Probably around schemas which guarantee their own validity and provide relevant operations in a centralized way
  • I/O especially would benefit
  • This would also make future refactoring of keypoints handling easier
• All I/O still heavily depends on pandas and its operations
• Relax versions pins (napari 0.7, ...)

Miscellaneous

• pydantic schemas in config/models.py should be moved to core
• Plugin metadata in layers is not namespaced, instead of a flat dict, using a "napari-dlc" key and adding all metadata there would make interoperability with other plugin/metadata edits safer
• Coverage report is biased by tests being included currently