[JasMehta08]

What

Adds the integration test harness from the test strategy (#31): a satchel-it crate whose tests drive the running Compose cluster over HTTP, a script that runs them locally, and a CI job that runs them on every PR.

Closes #32

Why

Week 3 build issues need somewhere to put tests that cross services, and a PR that breaks a service should fail CI. This mirrors Magic Pocket, where releases go through integration tests with failure injection before they move on [B3].

• Where: crates/satchel-it. The config helper is in src/lib.rs, the tests are in tests/cluster.rs.
• Kept out of cargo test --workspace: the tests are #[ignore]d and run with cargo test -p satchel-it -- --ignored. If the cluster isn't there they fail with "connection refused" for each service, they never pass silently.
• HTTP client: ureq with default features off (plain HTTP, no TLS, no OpenSSL). It's a dev-dependency, so the Docker image build doesn't compile it. Pinned to 3.4.2 in the lock file.
• Which services: every Satchel service in docker-compose.yml, read from docker compose config, so nothing needs changing when docs(design): finalize the exercise plan with RS(6,3) on 9 nodes #56's 9-node cluster lands. A declared service that's down fails the tests instead of being skipped.
• Config: each URL defaults to the service's Compose host port, and SATCHEL_IT_<SERVICE>_URL overrides it (for example SATCHEL_IT_NODE_Z1_URL), so the same suite can point at the AWS cluster later. Readiness polls /health for up to 60 s (SATCHEL_IT_READY_TIMEOUT_SECS), with no fixed sleeps.
• First tests: ts_svc_1_all_services_healthy and ts_node_4_nodes_report_their_zone. A failure lists every broken service with its URL and what it returned.
• Compose race from Set up local dev environment and verify the repo builds #1: the script and the CI job build satchel:dev once, then run docker compose up -d --wait without --build. docker-compose.yml itself is unchanged.

How to check it

cargo test --workspace            # the two integration tests show as ignored
scripts/integration-test.sh       # builds, starts the cluster, runs them, tears down

• The new Integration tests (Compose) job should pass on this PR.
• To see it fail: change "satchel-api" to anything else in crates/satchel-api/src/main.rs and run the script. It fails with:
  api at http://localhost:8080/health: expected service "satchel-api" with status "ok", got {"service":"satchel-broken",...}

Once the job has passed on main a few times, I'll ask @frectrium to make it a required check.

Checklist

• cargo fmt, cargo clippy and cargo test pass locally (code PRs)
• Tests added or updated for the change (code PRs)
• New sources are added to docs/sources/README.md (research PRs)
• Design decisions are recorded in the Design Decision Log
• Docs and README are updated if behaviour or setup changed