Add WYSIWYG capture preview with raw, processed, and masked scan views

[nikolareljin]

ID: KF-012

During capture, the GUI preview currently shows only a raw color
frame. The user cannot tell what is actually being saved, which surfaces are
inside the depth mask, whether ROI is excluding background correctly, or whether
HSV masking is isolating the target object. This makes scanning feel blind and
prevents the user from trusting the result.

Fix: replace the current preview with a WYSIWYG capture monitor that shows the
same processed data that will be written to disk. The preview should support:

• raw RGB view
• processed RGB view after depth/ROI/color masking
• depth view with color map
• combined side-by-side view as default

Overlay explicit indicators:

• active depth min/max
• ROI rectangle
• color mask enabled/disabled
• non-zero pixel ratio after masking
• warning badge when too much of the frame is background or too little of the
  object remains

Use the exact same preprocessing path as capture persistence so the preview is
truthful. Do not maintain a separate preview-only masking implementation.

Files: src/kinect_forge/capture.py, src/kinect_forge/gui.py,
docs/CAPTURE_WORKFLOW.md, docs/GUI.md

References:

• OpenCV colormaps: https://docs.opencv.org/4.x/d3/d50/group__imgproc__colormap.html
• Open3D RGB-D concepts: https://www.open3d.org/docs/latest/tutorial/geometry/rgbd_image.html