Ship a five-minute LiDAR-camera evidence demo

[rsasaki0109]

What changed

• Package a deterministic KITTI-shaped LiDAR-camera fixture in the wheel so the evidence demo works without ROS or a dataset download.
• Make calibrex demo kitti-lidar-camera-evidence --strict-assessment use that packaged fixture by default and resolve configured dataset paths relative to the selected config.
• Strengthen the fixture to detect 16/24 mandatory known-bad perturbations and pass all six falsification-policy gates without relaxing thresholds.
• Add generator-parity, integration, schema, and isolated-wheel release checks.
• Refocus the README and public-dataset tutorial on a reproducible five-minute path, with an explicit synthetic-fixture caveat.
• Prepare the accumulated post-v0.4.0 changes as version 0.4.1 and keep generated schemas and action references version-aligned.

Why

The first LiDAR-camera experience previously depended on a repository checkout and a fixture whose known-bad controls were inconclusive. A PyPI user should be able to run one honest, falsifiable evidence workflow immediately and verify its provenance.

User impact

After the 0.4.1 package is published, a new user can install Calibrex, run the strict LiDAR-camera demo in under one minute on the clean Windows wheel smoke path, inspect the HTML evidence, and verify the digest-bound bundle. The bundled data validates the pipeline and evidence contracts; it does not claim physical-rig calibration accuracy.

Validation

• python -m pytest: 762 passed, 16 skipped
• Post-version/schema focused suite: 56 passed, 1 skipped
• python -m ruff check .: passed
• python -m mypy src/calibrex: passed (190 source files)
• python -m mkdocs build --strict: passed
• python tools/local_release_smoke.py ...: passed for calibrex-0.4.1-py3-none-any.whl, including installed-wheel strict demo and raw-recomputed bundle verification