Add Docker support for LTEngine

[denyncrawford]

Summary

• Add Dockerfile, .dockerignore, docker-compose.yml, docs/README_DOCKER.md, docker/Dockerfile.cuda, and a CI workflow to enable building the LTEngine image.
• Multi-stage Dockerfile builds the release binary and produces a minimal runtime image.
• docker-compose.yml provides a quick way to run LTEngine mounting a models directory.

How to test

1. Build:
   docker build -t libretranslate/ltengine:local .
2. Run:
   docker run --rm -p 8080:8080 -v $(pwd)/models:/models -e MODEL_PATH=/models libretranslate/ltengine:local
3. Verify the service responds at the expected HTTP endpoints.

Notes / Caveats

• The Dockerfile expects the crate binary to be named ltengine. If the actual binary name differs, I can update the Dockerfile — tell me the final binary name or the Cargo.toml path.
• GPU/CUDA Dockerfile is a sketch; enable only if LTEngine includes CUDA support and the proper build flags.
• Models are intentionally not included in the image; mount them as a volume.

Follow-ups