feat(admin): add service control and site replication commands

[overtrue]

What

Adds the two admin command families that rc was missing relative to mc, as reported in rustfs/backlog#1156 (comment 4968601053):

• rc admin service <restart|stop|freeze|unfreeze> <ALIAS> — maps to POST /rustfs/admin/v3/service?action=<action>. restart and stop trigger a graceful shutdown (the server response explicitly states that a process manager is responsible for relaunch after restart); freeze/unfreeze pass through the server's advisory effective: false flag honestly.
• rc admin replicate <add|info|status|remove> <ALIAS>... — maps to /rustfs/admin/v3/site-replication/{add,info,status,remove}. Peer sites for add are given as configured alias names and resolved to endpoint + credentials from the local alias store, mirroring mc admin replicate add ergonomics. status supports --buckets/--users/--groups/--policies/--metrics section flags and defaults to the four summary sections like mc.

Wire types (PeerSiteSpec, SiteRemoveSpec, ServiceActionResult) mirror the server's crates/madmin definitions, including the MinIO-compat endpoints field name quirk on PeerSite. The native /rustfs/admin prefix accepts plain JSON, so everything reuses the existing SigV4 AdminClient request path — no madmin payload encryption needed.

Why

The rustfs/backlog#1156 analysis found operators must fall back to mc for site replication and to SSH + systemctl for service control. Notably, mc admin service stop is broken against RustFS regardless (mc streaming-signs the POST without content-length, which s3s rejects at signature.rs:445 before any handler runs), so a native rc command is the only clean client-side path for service control.

Additive only: no existing command, output schema, or exit code changes. The docs/reference/rc/admin.md contract gains new sections for the new commands.

Verification

• cargo fmt --all clean, cargo clippy --workspace --all-targets -- -D warnings zero warnings, cargo test --workspace all green (13 new integration tests in admin_service.rs / admin_replicate.rs following the existing admin_support mock-server pattern).
• Live end-to-end against two rustfs/rustfs main (ea2e24ac1) instances (with RUSTFS_REPLICATION_ALLOW_LOOPBACK_TARGET=true for loopback endpoints):
  • rc admin replicate add site1 site2 → cluster established; buckets, objects, and IAM users created on site1 replicated to site2 (verified via rc cat site2/... and rc admin user list site2).
  • rc admin replicate status site1 → sites + replicated entity counts; remove --all dissolves the cluster; state survives server restarts.
  • rc admin service stop/restart → server logs RustFS server stopped and the process exits gracefully; freeze/unfreeze report advisory status.
  • Interop: mc admin replicate info (MinIO compat path) reads the cluster created by rc.

Follow-ups (not in this PR)

• rc admin replicate edit / resync / repair subcommands (server endpoints already exist).
• Typed response structs for info/status instead of serde_json::Value passthrough, once the output shape is settled.

Closes the rc gap identified in rustfs/backlog#1156.

---

BREAKING CHANGE: none in behavior — this marker is present because docs/reference/rc/admin.md (a protected CLI contract file) gains new sections for the added service and replicate commands. The contract change is purely additive: no existing command, option, output schema, or exit code is modified.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.