chore: add visualizer service to docker-compose

[Colinho22]

Motivation

Once the visualizer (issue #??) and the dockerized runner (issue #??) both
exist, docker compose up visualizer should be the canonical way for a
replicator to open the results dashboard against an experiment DB. This
closes the replication loop: clone → build → run experiments → view
results, all in containers.

Proposal

Add a visualizer service to docker-compose.yml:

• Reuses the existing image (same dependencies), overrides command with streamlit run src/maestro/viz/app.py --server.port 8501 --server.address 0.0.0.0.
• Port 8501:8501 mapped to host.
• Read-only mount of the same DB the runner writes to.
• Does not require API keys (pure read path) so it can run without .env populated.

Update README:

• docker compose up visualizer → open http://localhost:8501.
• Note the read-only DB mount (no risk of corrupting experiment data).

Scope

• In scope:
  • Compose service definition
  • Port and volume wiring
  • README update
• Out of scope:
  • Separate image for the visualizer (reuse runner image until it diverges)
  • Hosted / public deployment
  • HTTPS / auth

Open questions

• Bundle visualizer deps (streamlit, plotly or altair) into the main image, or split into an optional extra (pip install -e ".[viz]")? Extra is cleaner but adds a build matrix.

Depends on

• Issue feat: results visualizer for the experiment #19
• Issue chore: finalize Docker setup for experiment runner #20

Related

• docker-compose.yml