Onboarding Workbench

Onboarding Workbench replaces the spreadsheet most platforms run alongside their console. Each tenant gets a `SysOnboardingChecklist` with eight default steps (contract, kickoff, configure, import data, train admins, dry-run, go/no-go, go-live), each carrying status, assignee, and a free-text blocking reason. `PATCH /sys/onboarding/{tenant_id}/steps/{step_id}` auto-flips status as fields update and stamps `completed_at` on the parent checklist when all steps pass; once that flag flips, the dashboard moves the tenant out of `active` into `completed`. The dashboard at `GET /sys/onboarding` lists all in-flight onboardings with an active/completed filter, days-open, current step, and assignee — the place a CSM lands every morning.

The data-import wizard at `POST /sys/onboarding/{tenant_id}/imports` queues a `SysOnboardingImport` for one of five formats (csv, sie, fipjp, xml, json), surfaces a validation preview with row counts and parse warnings, and tracks progress through to completion; it wraps the PL-T056 importer so the same code path serves both onboarding and ongoing imports. The tenant-health check at `POST /sys/onboarding/{tenant_id}/health-check` runs three pluggable probes registered through `set_health_probes()` — sample login, sample transaction, sample member report — and aggregates to pass/warn/fail. The aggregate is what unblocks the go/no-go step.

The onboarding retrospective at `GET /sys/onboarding/retrospective?group_by=quarter` slices completed onboardings by time-to-live, total days, and step-by-step duration; it lets the team prove whether onboarding is getting faster and identify which step keeps slipping (it is almost always `import data`). Together these surfaces collapse a multi-tool onboarding workflow into one console with audit, accountability, and trend data.