compose: start validator clients once charon is live

[KaloyanTanev]

Validator clients now start once charon is healthy instead of waiting for proposer-config.json to exist. On an existing node the file is already there, so a recreated VC rendered it right away, before charon (depends_on only waited for its container to start) had rewritten it with its current flags.

Charon's healthcheck moves from /readyz to /livez, and every VC (including the default Lodestar one) depends on charon with condition: service_healthy. Charon serves /livez only once started, after it has written or updated the file, so a healthy charon means the file is current. /readyz can't be used: it stays unready until a VC connects. start_interval: 1s keeps the wait short instead of the 30s default interval (Docker Engine 25+). The wrappers drop their wait loops and keep the fallback for a missing file. The VC test override gives its stub charon an always-healthy check.

Trade-off: docker ps shows charon healthy once it runs, not once it's ready; readiness stays in the app_monitoring_readyz metric and its alerts.

Tested with docker compose config on the real and test compose files, and with stub services: a VC depending on a charon stub that writes the file and serves /livez after 5s started in the same second the file was written. Pairs with ObolNetwork/charon#4784, which updates the builder settings of an existing proposer config on start.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud