refactor: rename orchestrator/contributor naming to router/operator

[bagelface]

Summary

• Renames the last leftover Orchestrator/Contributor naming from the deleted star-topology system (CLI flags, struct/fn/var names, config filenames) to the current router/operator terminology used everywhere else post-feat: integrate commonware-consensus aggregation engine #184
• Deletes config/orchestrator.json, an unreferenced duplicate of router_key.json (formerly router_orchestrator.json)
• Removes dead CONTRIBUTOR_1/2/3_KEYFILE entries from example.env (unused since nodes now get keys via direct file mounts, not env-var indirection)
• Rewrites scripts/README.md, which still described a pre-monorepo workflow (submodules, wrong ports, wrong CI file/branches, wrong package name)

Test plan

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets -- -D warnings
• cargo check --workspace --all-targets
• cargo test --workspace (56 passed, 0 failed)
• Repo-wide grep for orchestrator/contributor returns no hits outside intentional identifiers

[BreadrichEngels]

Clean post-#184 cleanup — the orchestrator/contributor → router/operator rename is applied consistently across the coupled surfaces, which is the thing that matters for a rename like this. LGTM.

Traced the couplings that would actually break if a rename were half-applied, and they all move together:

• File renames are consistent end to end. public_orchestrator.json → public_router.json (node volume mounts + the --router arg + node/router READMEs) and router_orchestrator.json → router_key.json (router volume + --key-file + the local-integration CI invocation + router README). The renamed files are similarity index 100%, so content is untouched and the RouterConfig struct still parses the same g2_*/address/port fields (serde keys unchanged).
• submitter.rs is a pure variable rename (contributor → operator) in the bitmap→operator resolution loop — no logic change to the block-pinned-read path.
• Removing the CONTRIBUTOR_*_KEYFILE env vars is safe — compose mounts each testacc*.private.bls.key.json to /app/key.json directly, so those vars were dead. And the scripts/README rewrite corrects real staleness (drops the submodule step, fixes ports to 3001-3003/4000, the commonware-avs-scripts package name, and the two workflow files).

One thing worth validating (the unit tests can't catch it): the --orchestrator → --router change is a CLI-interface rename between docker-compose.yml and the node binary. Your cargo test covers the code, but a compose↔binary flag mismatch only shows up at runtime — and local-integration-test.yml "pulls prebuilt node images." If that image is a stale tag built before this commit, it still expects --orchestrator and clap will reject the new --router flag, failing the run. Please confirm the node image the compose e2e pulls is built from this branch (or run the docker-compose / local-integration e2e once) so the flag rename is exercised end-to-end, not just the unit suite.

Minor: deleting config/orchestrator.json is fine per your repo-wide grep, but an in-repo grep can't see external deploy/ops scripts — just worth a glance that nothing outside the repo referenced that path (it's only a local test key, so low stakes either way).

CI isn't reporting on the head commit yet — confirm fmt/clippy/check/the 56 tests are green. Approving.