Add capability-aware platform UX and cross-platform reconstruct-only mode

[nikolareljin]

ID: KF-015

The app currently presents itself as a GUI that runs on Linux,
Windows, and macOS, but Kinect v1 capture is only realistic on Linux with
libfreenect. On non-Linux systems, users can still reconstruct existing scans,
yet the UI does not clearly separate "capture capable" from "reconstruct only".
This makes the app feel broken or misleading off Linux.

Fix: add startup capability detection and gate the UI by actual runtime support.
At launch, detect:

• operating system
• freenect import availability
• Kinect backend availability
• browser/viewer capability

Then present a capability matrix in the Status/Pipeline area:

• Live Kinect capture available
• Reconstruction available
• Viewer available
• Export available

On unsupported platforms, hide or disable capture-first actions and instead make
"Import Existing Scan" the primary CTA. Ensure docs match this behavior with
platform-specific quick starts. This is required for a production-grade,
user-friendly app that can still be useful on any platform even when live Kinect
capture is not supported there.

Files: src/kinect_forge/gui.py, src/kinect_forge/cli.py, docs/SETUP.md,
docs/USER_GUIDE.md, README.md

References:

• Tkinter cross-platform note already present locally
• Open3D platform support overview: https://www.open3d.org/docs/latest/getting_started.html
• libfreenect project: https://github.com/OpenKinect/libfreenect