Server-side heatmap rendering, dark theme redesign, uv-based Docker build

[Friedjof]

Summary

• Replace the slow embedded Grafana <iframe> on the status page with a server-side rendered "Opening Times" heatmap table. The data is fetched from Grafana's public dashboard query API, cached, and refreshed in a background thread so page loads never wait on the external call.
• Redesign the status page with a dark theme, a bigger/squared "Open/Closed" status card, and a smaller navbar.
• Migrate dependency management from requirements.txt to pyproject.toml + uv (with uv.lock for reproducible installs).
• Switch the Docker image build to install dependencies via uv instead of pip, and add a compose.yml for local development.
• Add CI workflows: lint.yml (ruff + hadolint on every push/PR) and build-test.yml (verifies the image still builds on main); docker-build-push.yml now only publishes to GHCR on version-tag pushes (v*) instead of every push to main.

Why

The Grafana iframe made the page noticeably slow to load just to show a small 7x24 probability table. Fetching and caching that data server-side removes the dependency on Grafana's frontend entirely for end users, while the background refresh keeps the data reasonably fresh without ever blocking a request.

Test plan

• docker compose up --build builds and starts the container; / returns 200 and renders the heatmap table.
• /simple.php, /api/v2/status.json etc. still respond correctly.
• Verified the heatmap falls back to a plain message if the Grafana request fails, instead of a 500 error.
• Verified pip/uv install works from a clean environment.
• Manual visual check of the new dark theme in a browser (recommended before merge).