fix: refresh stopped imported container names safely

[chen21019]

Root cause and bounded correction

The normal stop process retains an inactive instance/host mapping. The imported-container name-refresh predicate in 0.183.330 rejected that stable stopped/inactive pair, leaving old names on retained rollback containers.

Allow exactly running/active or stopped/inactive in the shared selection and name-only full-row CAS predicate. Preserve all existing account, active Host/Agent, full Docker ID, unique nonremoved mapping and managed/service/stack exclusions. No authentication/schema/migration or lifecycle-process change.

Evidence

• Ten distinct DAO cases passed across focused incremental runs. The official candidate workflow must retain all ten DAO, nine handler, three monitor and six role-schema cases.
• 84 synthetic CTE SELECT cases generated by the actual production predicate passed on MariaDB; no real platform tables or SQL mutations. This is not a real CAS-race or browser claim.
• Initial official CI at 1f03d1d failed before tests because the source/artifact gates still expected 330. Three expected-version/log lines are now 331; no gate was relaxed.
• Corrected candidate: 515a5d3. Exact-source build, artifact scan, CodeQL and independent review are required before publication/merge.

Prepare numeric immutable 0.183.331 only. Consuming Server500 and actual native browser name verification remain pending; no existing tag or image will be overwritten.