Reproducible environment with Docker

[Sammyjoseph999]

Adds a Docker path so any user can reproduce an identical runtime environment, independent of their local Python version.

What

• Dockerfile — multi-stage, uv-based, non-root, CPython 3.10 (matches CI). Reproducibility comes from three pins: base image pinned by digest, uv pinned to an exact release (0.11.28), and uv.lock (every dependency fixed to exact version + hash). Dependency layer is cached separately from source for fast rebuilds.
• .dockerignore — keeps the build context small/deterministic (installs from pyproject.toml + uv.lock).
• docker-compose.yml — wraps the Earth Engine credential mount, GCP_PROJECT_ID, and outputs/ volume.
• README — new "Reproducible environment with Docker" section.

Design notes

• Earth Engine credentials are not baked into the image (per-user OAuth). They mount read-only at /home/app/.config/earthengine at runtime; GCP_PROJECT_ID is a runtime env var. The offline surface (import, --help, unit suite, NASA POWER) needs neither.
• Base image pinned by digest so the OS layer is reproducible over time, not just today.
• Respects existing runtime conventions (env-var config, outputs/cache/...).

Verification (built and run locally)

• docker build succeeds with no compiler (all deps resolve to manylinux cp310 wheels).
• docker run --rm climate-toolkit --help lists all subcommands.
• Import + version check passes inside the container.
• Container runs as non-root (uid 1000).