Replace unwrap() with ? in control_db for proper error propagation

[Shiven0504]

Summary

• Replace 5 unwrap() calls in control_db.rs with ? to propagate errors instead of panicking at runtime (bsatn::to_vec, bsatn::from_slice, and get_replicas)
• Change get_leader_replica_by_database return type from Option<Replica> to Result<Option<Replica>> across the ControlStateDelegate trait and all implementations/call sites
• Fix help text for --listen-addr: duplicate word ("to to") and incorrect default port (said 80, actual default is 3000)

Details

The unwrap() calls in control_db.rs could cause panics if serialization/deserialization ever failed. Since all the containing functions already return Result, replacing with ? is straightforward and lets callers handle errors gracefully.

The get_leader_replica_by_database method was the only replica-related method on the trait that didn't return Result, despite calling get_replicas() which can fail. This inconsistency was masked by the unwrap().

Files changed

File	Change
crates/standalone/src/control_db.rs	Replace 5 unwrap() → ?, change return type
crates/client-api/src/lib.rs	Update trait definition + blanket impl
crates/standalone/src/lib.rs	Update trait impl + call site
crates/client-api/src/routes/database.rs	Handle new Result at call site
crates/testing/src/modules.rs	Handle new Result in test helper
crates/standalone/src/subcommands/start.rs	Fix help text typo

Test plan

• CI compilation and tests should pass — no logic changes, only error propagation
• The get_leader_replica_by_database callers all handle the new Result appropriately