fix(cli): recover() in multi_service_engine_versions worker goroutine #997

Summary

cmd/multi_service_engine_versions.go:112 runs a wg.Wait()-joined goroutine that calls queryRDSInstancesInRegion (map writes + AWS SDK calls, both can panic) with no recover(). A panic there crashes the CLI process. This is the CLI batch path, out of scope for #672 (which covered the 3 Lambda fire-and-forget goroutines, fixed in PR #859), but it is the same class of defect.

Acceptance criteria

• Add a deferred recover() (logging the panic + region context) at the top of the goroutine in cmd/multi_service_engine_versions.go, mirroring the pattern PR chore(reliability): recover() in fire-and-forget goroutines (closes #672) #859 applied to the Lambda goroutines.
• A panic in one region's worker must not crash the whole run; it should log and let the other regions complete.