fix(schema): make current V2 the sole fresh-install DDL owner

[servicememeai]

Problem

Previous startup still ran historical cloudfile.sql and the 001..042 legacy SchemaRunner, creating Guard/ACL/group tables incompatible with current V2 (including the SQL 1176 Guard index outage).

Fix

• Fresh V2 bootstrap consumes Server's cloudfile-v2.sql (exactly cf_library_storage, cf_dir_acl, cf_sso_group_map), plus cloudfile-v2-edit.sql only if Checkout or file locks are explicitly enabled.
• Stop applying legacy cloudfile.sql, old hierarchy ALTER and old audit table DDL during normal V2 startup; retain the CE-native metadata/tag compatibility shims.
• Preserve .cloudfile-schema-init-pending intent on failure; existing volumes do not auto-upgrade.
• Add /scripts/cloudfile-v2-schema.py plan|check for metadata-only inspection of existing tables, no data reads or persistent mutations.
• Update targeted unit tests and document backup/offline migration and safe non-destructive recovery. Historical upgrade guide explicitly marked historical.

Dependencies

Hub #41 provides the V2 inventory. Server #20 enforces current Guard layout and stops deterministic upload retry loops.

Limits

No migration of existing Guard rows is attempted: preserve data/generation and active leases until a reviewed offline snapshot-specific migration. No DEV/PROD changes or full costly CI automatically.