Migrate from pip/setuptools to uv for dependency management

[halfguru]

Summary

This PR migrates msc-pygeoapi from pip/setuptools to uv for dependency management, following modern Python packaging standards (PEP 621).

Motivation

• Performance: uv is significantly faster, especially with cached packages
• Modern standards: pyproject.toml as single source of truth
• Developer experience: Cleaner dependency management, better resolution
• Industry adoption: Widely adopted by major projects (pytest, requests, flask)

Installation Time Comparison

Scenario	Time
pip (cold cache)	65s
pip (warm cache)	33s
uv (cold cache)	42s
uv (warm cache)	<1s

uv with warm cache is 33x faster than pip with warm cache.

Changes

• Add pyproject.toml with project metadata, dependencies, and dev dependencies
• Keep minimal setup.py for Debian packaging compatibility (debuild)
• Update CI workflow to use astral-sh/setup-uv@v7
• Consolidate flake8 check into main workflow (3 parallel jobs: lint, test, debian)
• Update Dockerfile to use uv pip install --system and Docker buildx
• Update deploy/nightly/deploy-nightly.sh to use pip install . (modern standard)
• Update documentation (README.md, docker/README.md)
• Remove requirements.txt, requirements-dev.txt, requirements-oracle.txt

Dependency Versions

All 58 common dependencies resolve to identical versions. Minor additions in uv:

• coverage and pytest-cov added to dev dependencies for better test coverage

Backward Compatibility

• ✅ Debian packaging maintained via minimal setup.py
• ✅ All existing functionality preserved
• ✅ Same package versions (no breaking changes)
• ✅ Docker builds work identically

Tests Performed

• ✅ Unit tests: 9/9 passed
• ✅ Linting: No errors
• ✅ Docker build: Successful
• ✅ Docker buildx: Successful
• ✅ Dependency version comparison: Identical

Notes

• gdal is not included in pyproject.toml dependencies - it must be installed via system packages (apt install python3-gdal or libgdal-dev + pip) due to build requirements.