ci(rust): wait for CI Postgres readiness + retry integration tests on transient DB flake

[BryanFRD]

Problem

The integration job in reusable-ci-rust.yml flakes ~50% on main (FerrLabs-Cloud idor_scoping panics at Pool<Postgres> init). createdb on the shared postgres-ci pod succeeds, then the test's pool connect intermittently fails — the pod briefly refuses connections under concurrent CI load. The tests are deterministic; only the DB setup flakes.

Fix

1. Readiness wait after createdb — pg_isready loop until the fresh DB accepts a connection before handing it to the tests.
2. Retry the cargo test integration step up to 3× — a transient connection flake retries instead of failing the build; a genuine regression still fails all three attempts.

Both are contained to the integration job. Applies to every repo that consumes this reusable workflow (all product Rust APIs). YAML validated.

Closes #149.