db status reads the bootstrap flag from any tenant

After a clean compass db recreate -y -k and a full bootstrap setup, the application works — the tenant exists, staff sign in, trades list — and yet:

🔧  Bootstrap mode

    true  — system provisioning wizard not yet run

🏢  Tenants (non-system)

    (none)

The setting is per tenant, and the table holds a current row for each:

select tenant_id, value from ores_variability_system_settings_tbl
 where name = 'system.bootstrap_mode'
   and valid_to = ores_utility_infinity_timestamp_fn();

tenant	value
ffffffff-… (Root Tenant)	false
24da711c-… (a test tenant)	true
6ff24e8c-… (a test tenant)	true

compass_db.bootstrap_mode() asks for the first current row and names no tenant, so whichever row the planner returns decides. A test run that creates a tenant and never finishes its onboarding therefore makes a real installation report bootstrap mode. The tenant list beside it answered "(none)" while eight tenants existed, so the same query shape is wrong in more than one place.

Two ways to fix it, and the second is better: filter on the system tenant, whose id is well known; or make the query count rows rather than take the first, and say "partly bootstrapped" when tenants disagree, which is the truth a person running this command wants.

It matters because the status command is what a person reads before deciding to recreate a database — the destructive step this session took on its say-so.