Add live scan diagnostics and capture quality feedback before reconstruction

[nikolareljin]

ID: KF-013

Even with better preview, the user still does not know whether the
scan is "good enough" to reconstruct. The app does not show what coverage has
been captured, whether too many frames were skipped, whether background leakage
is contaminating the scan, or whether the object was seen from enough angles.
Users need a quality gate before spending time on reconstruction.

Fix: add live diagnostics during capture and a post-capture quality summary.
Minimum diagnostics:

• saved frame count vs requested frame count
• skipped frame count in turntable mode
• estimated turntable/angle coverage for saved frames
• masked foreground area trend over time
• background leakage score
• repeated-frame or low-motion warning

After capture completes, show a "Scan Quality" panel with one of:

• Ready to reconstruct
• Likely incomplete coverage
• Too much background detected
• Reframe object and recapture

Generate a contact sheet or sample-strip preview from saved frames so the user
can verify object isolation without opening the filesystem manually. The quality
summary should feed directly into the next-step CTA in KF-014.

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

References:

• Related workflow issue: Add per-frame progress callback to reconstruct_mesh #8
• Capture QA inspiration, COLMAP-style image inspection concepts:
  https://colmap.github.io/tutorial.html