V136: Employment history for phoenix_kit_staff.

Replaces the single flat employment span on phoenix_kit_staff_people (employment_type / employment_start_date / employment_end_date / job_title / work_location) with a first-class history of employment spans, surfaced as a dedicated tab on the person profile.

Creates phoenix_kit_staff_employments — one row per span, carrying the employment type, a translatable job_title, the org placement at the time (primary_department_uuid + a primary_team_uuid snapshot), the date range (employment_end_date IS NULL = the current/open span), work_location, and free-text notes.

Single open span per person

A partial unique index enforces at most one open span (employment_end_date IS NULL) per person — the "current" employment. The app context closes the prior open span when a new one starts.

Denormalized "current" mirror

The matching columns already on phoenix_kit_staff_people are kept as a denormalized mirror of the current (open) span — the app's sync_current/1 writes them in the same transaction as any span change, so existing readers (overview org tree, people list) need no join. This migration does NOT drop those columns.

Backfill

One open span per existing person is seeded from their current columns (employment_type / job_title / dates / primary_department_uuid / work_location), copying any per-locale job_title overrides out of the person's translations JSONB into the span's translations. Guarded by a NOT EXISTS check on the span table so a re-run is a safe no-op; people with no employment data at all are skipped (they start with an empty history). primary_team_uuid is left null on backfill — the person's team comes from the many-to-many team_memberships, which has no single "primary" to copy.